Typical	O
generative	O
model	O
approaches	O
include	O
naive	B-algorithm
Bayes	I-algorithm
classifier	I-algorithm
s	O
,	O
Gaussian	B-algorithm
mixture	I-algorithm
model	I-algorithm
s	O
,	O
variational	B-algorithm
autoencoders	I-algorithm
and	O
others	O
.	O

Finally	O
,	O
every	O
other	O
year	O
,	O
ELRA	B-conference
organizes	O
a	O
major	O
conference	O
LREC	B-conference
,	O
the	O
International	B-conference
Language	I-conference
Resources	I-conference
and	I-conference
Evaluation	I-conference
Conference	I-conference
.	O

The	O
task	O
is	O
usually	O
to	O
derive	O
the	O
maximum	B-algorithm
likelihood	I-algorithm
estimate	I-algorithm
of	O
the	O
parameters	O
of	O
the	O
HMM	B-algorithm
given	O
the	O
of	O
output	O
sequences	O
.	O

Unlike	O
neural	B-algorithm
network	I-algorithm
s	O
and	O
Support	B-algorithm
vector	I-algorithm
machine	I-algorithm
,	O
the	O
AdaBoost	B-algorithm
training	O
process	O
selects	O
only	O
those	O
features	O
known	O
to	O
improve	O
the	O
predictive	O
power	O
of	O
the	O
model	O
,	O
reducing	O
dimensionality	O
and	O
potentially	O
improving	O
execution	O
time	O
as	O
irrelevant	O
features	O
need	O
not	O
be	O
computed	O
.	O

Troponymy	B-misc
is	O
one	O
of	O
the	O
possible	O
relations	O
between	O
verb	O
s	O
in	O
the	O
semantic	B-misc
network	I-misc
of	O
the	O
WordNet	B-product
database	I-product
.	O

A	O
frame	O
language	O
is	O
a	O
technology	O
used	O
for	O
knowledge	B-task
representation	I-task
in	O
artificial	B-field
intelligence	I-field
.	O

NIST	B-metrics
also	O
differs	O
from	O
Bilingual	B-metrics
evaluation	I-metrics
understudy	I-metrics
in	O
its	O
calculation	O
of	O
the	O
brevity	B-misc
penalty	I-misc
insofar	O
as	O
small	O
variations	O
in	O
translation	O
length	O
do	O
not	O
impact	O
the	O
overall	O
score	O
as	O
much	O
.	O

The	O
model	O
is	O
initially	O
fit	O
on	O
a	O
training	O
dataset	O
,	O
The	O
model	O
(	O
e.g.	O
a	O
neural	B-algorithm
net	I-algorithm
or	O
a	O
naive	B-algorithm
Bayes	I-algorithm
classifier	I-algorithm
)	O
is	O
trained	O
on	O
the	O
training	O
dataset	O
using	O
a	O
supervised	B-field
learning	I-field
method	O
,	O
for	O
example	O
using	O
optimization	O
methods	O
such	O
as	O
gradient	B-algorithm
descent	I-algorithm
or	O
stochastic	B-algorithm
gradient	I-algorithm
descent	I-algorithm
.	O

FrameNet	B-product
has	O
been	O
used	O
in	O
applications	O
like	O
question	B-task
answering	I-task
,	O
paraphrasing	B-task
,	O
recognizing	B-task
textual	I-task
entailment	I-task
,	O
and	O
information	B-task
extraction	I-task
,	O
either	O
directly	O
or	O
by	O
means	O
of	O
Semantic	B-task
Role	I-task
Labeling	I-task
tools	O
.	O

This	O
would	O
include	O
programs	O
such	O
as	O
data	B-field
analysis	I-field
and	O
extraction	O
tools	O
,	O
spreadsheets	B-misc
(	O
e.g.	O
Excel	B-product
)	O
,	O
databases	B-misc
(	O
e.g.	O
Access	B-product
)	O
,	O
statistical	B-field
analysis	I-field
(	O
e.g.	O
SAS	B-product
)	O
,	O
generalized	B-misc
audit	I-misc
software	I-misc
(	O
e.g.	O
ACL	B-product
,	O
Arbutus	B-product
,	O
EAS	B-product
)	O
,	O
business	B-misc
intelligence	I-misc
(	O
e.g.	O
Crystal	B-product
Reports	I-product
and	O
Business	B-product
Objects	I-product
)	O
,	O
etc	O
.	O

Rethink	B-organisation
Robotics	I-organisation
-	O
founded	O
by	O
Rodney	B-researcher
Brooks	I-researcher
,	O
previously	O
with	O
iRobot	B-organisation
-	O
introduced	O
Baxter	B-product
in	O
September	O
2012	O
;	O
as	O
an	O
industrial	B-product
robot	I-product
designed	O
to	O
safely	O
interact	O
with	O
neighboring	O
human	O
workers	O
,	O
and	O
be	O
programmable	O
for	O
performing	O
simple	O
tasks	O
.	O

Typical	O
text	B-field
mining	I-field
tasks	O
include	O
text	B-task
categorization	I-task
,	O
text	B-task
clustering	I-task
,	O
concept	B-task
/	I-task
entity	I-task
extraction	I-task
,	O
production	B-task
of	I-task
granular	I-task
taxonomies	I-task
,	O
sentiment	B-task
analysis	I-task
,	O
document	B-task
summarization	I-task
,	O
and	O
entity	B-task
relation	I-task
modeling	I-task
(	O
i.e.	O
,	O
learning	O
relations	O
between	O
named	B-task
entity	I-task
recognition	I-task
)	O
.	O

Nonetheless	O
,	O
stemming	O
reduces	O
precision	B-metrics
,	O
or	O
TRUE	B-metrics
negative	I-metrics
rate	I-metrics
,	O
for	O
such	O
systems	O
.	O

A	O
special	O
case	O
of	O
keyword	B-task
spotting	I-task
is	O
wake	B-misc
word	I-misc
(	O
also	O
called	O
hot	B-misc
word	I-misc
)	O
detection	O
used	O
by	O
personal	O
digital	O
assistants	O
such	O
as	O
Alexa	B-product
or	O
Siri	B-product
to	O
wake	O
up	O
when	O
their	O
name	O
is	O
spoken	O
.	O

Prova	B-programlang
is	O
an	O
open	O
source	O
programming	O
language	O
that	O
combines	O
Prolog	B-programlang
with	O
Java	B-programlang
.	O

In	O
1987	O
,	O
Tocibai	B-organisation
Machine	I-organisation
,	O
a	O
subsidiary	O
of	O
Toshiba	B-organisation
,	O
was	O
accused	O
of	O
illegally	O
selling	O
CNC	B-product
milling	I-product
s	O
used	O
to	O
produce	O
very	O
quiet	O
submarine	O
propellers	O
to	O
the	O
Soviet	B-country
Union	I-country
in	O
violation	O
of	O
the	O
CoCom	B-organisation
agreement	O
,	O
an	O
international	O
embargo	O
on	O
certain	O
countries	O
to	O
COMECON	B-misc
countries	O
.	O

Engelberger	B-researcher
's	O
most	O
famous	O
co-invention	O
,	O
the	O
Unimate	B-product
industrial	I-product
robotic	I-product
arm	I-product
,	O
was	O
among	O
the	O
first	O
inductees	O
into	O
the	O
Robot	B-location
Hall	I-location
of	I-location
Fame	I-location
in	O
2003	O
.	O

Originally	O
controlled	O
via	O
static	B-misc
html	I-misc
web	O
pages	O
using	O
CGI	B-misc
,	O
work	O
by	O
Dalton	B-person
saw	O
the	O
introduction	O
of	O
an	O
augmented	B-field
reality	I-field
Java	B-programlang
-based	O
interface	O
that	O
met	O
with	O
limited	O
success	O
.	O

The	O
first	O
publication	O
about	O
the	O
LMF	B-task
specification	I-task
as	O
it	O
has	O
been	O
ratified	O
by	O
ISO	B-organisation
(	O
this	O
paper	O
became	O
(	O
in	O
2015	O
)	O
the	O
9th	O
most	O
cited	O
paper	O
within	O
the	O
LREC	B-conference
conferences	O
from	O
LREC	B-conference
papers	O
)	O
:	O

A	O
confusion	B-metrics
matrix	I-metrics
or	O
matching	O
matrix	O
is	O
often	O
used	O
as	O
a	O
tool	O
to	O
validate	O
the	O
accuracy	B-metrics
of	O
k	B-algorithm
-NN	I-algorithm
classification	I-algorithm
.	O

Decision	B-algorithm
tree	I-algorithm
learning	O
is	O
one	O
of	O
the	O
predictive	O
modeling	O
approaches	O
used	O
in	O
statistics	B-field
,	O
data	B-field
mining	I-field
and	O
machine	B-field
learning	I-field
.	O

At	O
runtime	O
,	O
the	O
target	O
prosody	B-misc
of	O
a	O
sentence	O
is	O
superimposed	O
on	O
these	O
minimal	O
units	O
by	O
means	O
of	O
signal	B-field
processing	I-field
techniques	O
such	O
as	O
linear	B-algorithm
predictive	I-algorithm
coding	I-algorithm
,	O
PSOLA	B-algorithm

This	O
approach	O
utilized	O
artificial	B-field
intelligence	I-field
and	O
machine	B-field
learning	I-field
to	O
allow	O
researchers	O
to	O
visibly	O
compare	O
conventional	O
and	O
thermal	O
facial	B-task
imagery	I-task
.	O

In	O
computer	B-field
science	I-field
,	O
evolutionary	B-algorithm
computation	I-algorithm
is	O
a	O
family	O
of	O
algorithms	O
for	O
global	B-task
optimization	I-task
inspired	O
by	O
biological	B-misc
evolution	I-misc
,	O
and	O
the	O
subfield	O
of	O
artificial	B-field
intelligence	I-field
and	O
soft	B-field
computing	I-field
studying	O
these	O
algorithms	O
.	O

For	O
instance	O
,	O
one	O
can	O
combine	O
some	O
measure	O
based	O
on	O
the	O
confusion	B-metrics
matrix	I-metrics
with	O
the	O
mean	B-metrics
squared	I-metrics
error	I-metrics
evaluated	O
between	O
the	O
raw	O
model	O
outputs	O
and	O
the	O
actual	O
values	O
.	O

The	O
majority	O
are	O
results	O
of	O
the	O
word2vec	B-product
model	I-product
developed	O
by	O
Mikolov	B-researcher
et	O
al	O
or	O
variants	O
of	O
word2vec	B-product
.	O

It	O
was	O
during	O
this	O
time	O
that	O
a	O
total	O
of	O
43	O
publications	O
were	O
recognized	O
by	O
the	O
CVPR	B-conference
and	O
the	O
International	B-conference
Conference	I-conference
on	I-conference
Computer	I-conference
Vision	I-conference
(	O
ICCV	B-conference
)	O
.	O

The	O
AIBO	B-product
has	O
seen	O
much	O
use	O
as	O
an	O
inexpensive	O
platform	O
for	O
artificial	B-field
intelligence	I-field
education	O
and	O
research	O
,	O
because	O
integrates	O
a	O
computer	O
,	O
Computer	B-field
vision	I-field
,	O
and	O
articulators	O
in	O
a	O
package	O
vastly	O
cheaper	O
than	O
conventional	O
research	O
robots	O
.	O

She	O
served	O
as	O
Program	O
Chair	O
of	O
International	B-conference
Conference	I-conference
on	I-conference
Computer	I-conference
Vision	I-conference
2021	I-conference
.	O

Scheinman	B-researcher
,	O
after	O
receiving	O
a	O
fellowship	O
from	O
Unimation	B-organisation
to	O
develop	O
his	O
designs	O
,	O
sold	O
those	O
designs	O
to	O
Unimation	B-organisation
who	O
further	O
developed	O
them	O
with	O
support	O
from	O
General	B-organisation
Motors	I-organisation
and	O
later	O
marketed	O
it	O
as	O
the	O
Programmable	B-product
Universal	I-product
Machine	I-product
for	I-product
Assembly	I-product
(	O
PUMA	B-product
)	O
.	O

An	O
overview	O
of	O
calibration	O
methods	O
for	O
binary	B-task
classification	I-task
and	O
multiclass	B-task
classification	I-task
classification	I-task
tasks	I-task
is	O
given	O
by	O
Gebel	B-researcher
(	O
2009	O
)	O

He	O
is	O
involved	O
in	O
fields	O
such	O
as	O
optical	B-task
character	I-task
recognition	I-task
(	O
OCR	B-task
)	O
,	O
speech	B-task
synthesis	I-task
,	O
speech	B-task
recognition	I-task
technology	O
,	O
and	O
electronic	O
keyboard	O
instruments	O
.	O

For	O
more	O
recent	O
and	O
state-of-the-art	O
techniques	O
,	O
Kaldi	B-product
toolkit	I-product
can	O
be	O
used	O
.	O

Johnson-Laird	B-researcher
is	O
a	O
Fellow	O
of	O
the	O
American	B-organisation
Philosophical	I-organisation
Society	I-organisation
,	O
a	O
Fellow	O
of	O
the	O
Royal	B-organisation
Society	I-organisation
,	O
a	O
Fellow	O
of	O
the	O
British	B-organisation
Academy	I-organisation
,	O
a	O
William	B-researcher
James	I-researcher
Fellow	O
of	O
the	O
Association	B-organisation
for	I-organisation
Psychological	I-organisation
Science	I-organisation
,	O
and	O
a	O
Fellow	O
of	O
the	O
Cognitive	B-organisation
Science	I-organisation
Society	I-organisation
.	O

At	O
the	O
IEEE	B-conference
International	I-conference
Conference	I-conference
on	I-conference
Image	I-conference
Processing	I-conference
in	O
2010	O
,	O
Rui	B-researcher
Hu	I-researcher
,	O
Mark	B-researcher
Banard	I-researcher
,	O
and	O
John	B-researcher
Collomosse	I-researcher
extended	O
the	O
HOG	B-algorithm
descriptor	I-algorithm
for	O
use	O
in	O
sketch	B-task
based	I-task
image	I-task
retrieval	I-task
(	O
SBIR	B-task
)	O
.	O

BLEU	B-metrics
uses	O
a	O
modified	O
form	O
of	O
precision	B-metrics
to	O
compare	O
a	O
candidate	O
translation	O
against	O
multiple	O
reference	O
translations	O
.	O

For	O
the	O
case	O
of	O
a	O
general	O
base	O
space	O
math	O
(	O
Y	O
,	O
\	O
mathcal	O
{	O
B	O
}	O
,	O
\	O
nu	O
)	O
/	O
math	O
(	O
i.e.	O
a	O
base	O
space	O
which	O
is	O
not	O
countable	O
)	O
,	O
one	O
typically	O
considers	O
the	O
relative	B-metrics
entropy	I-metrics
.	O

As	O
of	O
October	O
2011	O
,	O
the	O
already-existing	O
partnerships	O
with	O
the	O
United	B-country
States	I-country
'	O
National	B-organisation
Park	I-organisation
Service	I-organisation
(	O
NPS	B-organisation
)	O
,	O
the	O
United	B-country
Kingdom	I-country
's	O
Historic	B-organisation
Scotland	I-organisation
(	O
HS	B-organisation
)	O
,	O
World	B-organisation
Monuments	I-organisation
Fund	I-organisation
,	O
and	O
Mexico	B-country
's	O
Instituto	B-organisation
Nacional	I-organisation
de	I-organisation
Antropología	I-organisation
y	I-organisation
Historia	I-organisation
(	O
INAH	B-organisation
)	O
had	O
been	O
greatly	O
expanded	O
,	O
,	O
CyArk	B-misc
website	B-misc

Kernel	B-algorithm
SVMs	I-algorithm
are	O
available	O
in	O
many	O
machine-learning	B-field
toolkits	O
,	O
including	O
LIBSVM	B-product
,	O
MATLAB	B-product
,	O
and	O
others	O
.	O

The	O
2009	O
Loebner	B-misc
Prize	I-misc
Competition	I-misc
was	O
held	O
September	O
6	O
,	O
2009	O
at	O
the	O
Brighton	B-location
Centre	I-location
,	O
Brighton	B-location
UK	B-country
in	O
conjunction	O
with	O
the	O
Interspeech	B-conference
2009	I-conference
conference	I-conference
.	O

The	O
humanoid	O
QRIO	B-product
robot	I-product
was	O
designed	O
as	O
the	O
successor	O
to	O
AIBO	B-product
,	O
and	O
runs	O
the	O
same	O
base	O
R-CODE	B-product
Aperios	B-product
operating	I-product
system	I-product
.	O

Speech	B-misc
waveforms	I-misc
are	O
generated	O
from	O
HMMs	B-algorithm
themselves	O
based	O
on	O
the	O
maximum	B-algorithm
likelihood	I-algorithm
criterion	O
.	O

Google	B-product
Translate	I-product
is	O
a	O
free	O
multilingual	B-task
statistical	I-task
machine	I-task
translation	I-task
and	O
neural	B-task
machine	I-task
translation	I-task
service	O
developed	O
by	O
Google	B-product
,	O
to	O
translate	O
text	O
and	O
websites	O
from	O
one	O
language	O
into	O
another	O
.	O

Skeletons	O
are	O
widely	O
used	O
in	O
computer	B-field
vision	I-field
,	O
image	B-field
analysis	I-field
,	O
pattern	B-field
recognition	I-field
and	O
digital	B-field
image	I-field
processing	I-field
for	O
purposes	O
such	O
as	O
optical	B-task
character	I-task
recognition	I-task
,	O
fingerprint	B-task
recognition	I-task
,	O
visual	B-task
inspection	I-task
or	I-task
compression	I-task
.	O

The	O
ImageNet	B-conference
Large	I-conference
Scale	I-conference
Visual	I-conference
Recognition	I-conference
Challenge	I-conference
is	O
a	O
benchmark	O
in	O
object	B-task
classification	I-task
and	I-task
detection	I-task
,	O
with	O
millions	O
of	O
images	O
and	O
hundreds	O
of	O
object	O
classes	O
.	O

Bengio	B-researcher
,	O
together	O
with	O
Geoffrey	B-researcher
Hinton	I-researcher
and	O
Yann	B-researcher
LeCun	I-researcher
,	O
are	O
referred	O
to	O
by	O
some	O
as	O
the	O
Godfathers	B-misc
of	I-misc
AI	I-misc
and	O
Godfathers	B-misc
of	I-misc
Deep	I-misc
Learning	I-misc
.	O

He	O
is	O
a	O
Life	O
Fellow	O
of	O
IEEE	B-organisation
.	O

NSA	B-organisation
Bethesda	I-organisation
is	O
responsible	O
for	O
base	O
operational	O
support	O
for	O
its	O
major	O
tenant	O
,	O
the	O
Walter	B-organisation
Reed	I-organisation
National	I-organisation
Military	I-organisation
Medical	I-organisation
Center	I-organisation
.	O

The	O
three	O
major	O
learning	O
paradigms	O
are	O
supervised	B-field
learning	I-field
,	O
unsupervised	B-field
learning	I-field
and	O
reinforcement	B-field
learning	I-field
.	O

Examples	O
include	O
control	B-task
,	O
planning	B-task
and	I-task
scheduling	I-task
,	O
the	O
ability	O
to	O
answer	B-task
diagnostic	I-task
and	I-task
consumer	I-task
questions	I-task
,	O
handwriting	B-task
recognition	I-task
,	O
natural	B-task
language	I-task
understanding	I-task
,	O
speech	B-task
recognition	I-task
and	O
facial	B-task
recognition	I-task
.	O

In	O
1991	O
he	O
was	O
elected	O
as	O
a	O
fellow	O
of	O
the	O
Association	B-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Artificial	I-conference
Intelligence	I-conference
(	O
1990	O
,	O
founding	O
fellow	O
)	O
.	O

However	O
,	O
by	O
formulating	O
the	O
problem	O
as	O
the	O
solution	O
of	O
a	O
Toeplitz	B-misc
matrix	I-misc
and	O
using	O
Levinson	B-algorithm
recursion	I-algorithm
,	O
we	O
can	O
relatively	O
quickly	O
estimate	O
a	O
filter	O
with	O
the	O
smallest	O
mean	B-metrics
squared	I-metrics
error	I-metrics
possible	O
.	O

In	O
July	O
2011	O
the	O
15th	B-conference
edition	I-conference
of	I-conference
Campus	I-conference
Party	I-conference
Spain	I-conference
will	O
be	O
held	O
at	O
the	O
City	B-location
of	I-location
Arts	I-location
and	I-location
Sciences	I-location
in	O
Valencia	B-location
.	O

Often	O
this	O
is	O
generally	O
only	O
possible	O
at	O
the	O
very	O
end	O
of	O
complicated	O
games	O
such	O
as	O
chess	B-product
or	O
go	B-product
,	O
since	O
it	O
is	O
not	O
computationally	O
feasible	O
to	O
look	O
ahead	O
as	O
far	O
as	O
the	O
completion	O
of	O
the	O
game	O
,	O
except	O
towards	O
the	O
end	O
,	O
and	O
instead	O
,	O
positions	O
are	O
given	O
finite	O
values	O
as	O
estimates	O
of	O
the	O
degree	O
of	O
belief	O
that	O
they	O
will	O
lead	O
to	O
a	O
win	O
for	O
one	O
player	O
or	O
another	O
.	O

The	O
difference	O
between	O
the	O
multinomial	B-algorithm
logit	I-algorithm
model	I-algorithm
and	O
numerous	O
other	O
methods	O
,	O
models	O
,	O
algorithms	O
,	O
etc.	O
with	O
the	O
same	O
basic	O
setup	O
(	O
the	O
perceptron	B-algorithm
algorithm	I-algorithm
,	O
support	B-algorithm
vector	I-algorithm
machine	I-algorithm
s	O
,	O
linear	B-algorithm
discriminant	I-algorithm
analysis	I-algorithm
,	O
etc	O
.	O

Association	B-conference
for	I-conference
Computational	I-conference
Linguistics	I-conference
,	O
published	O
by	O

In	O
computerised	O
Facial	B-product
recognition	I-product
system	I-product
,	O
each	O
face	O
is	O
represented	O
by	O
a	O
large	O
number	O
of	O
pixel	O
values	O
.	O

In	O
2002	O
,	O
his	O
son	O
,	O
Daniel	B-person
Pearl	I-person
,	O
a	O
journalist	O
working	O
for	O
the	O
Wall	B-organisation
Street	I-organisation
Journal	I-organisation
was	O
kidnapped	O
and	O
murdered	O
in	O
Pakistan	B-country
,	O
leading	O
Judea	B-person
and	O
the	O
other	O
members	O
of	O
the	O
family	O
and	O
friends	O
to	O
create	O
the	O
Daniel	B-organisation
Pearl	I-organisation
Foundation	I-organisation
.	O

As	O
of	O
late	O
2006	O
,	O
Red	B-organisation
Envelope	I-organisation
Entertainment	I-organisation
also	O
expanded	O
into	O
producing	O
original	O
content	O
with	O
filmmakers	O
such	O
as	O
John	B-person
Waters	I-person
.	O

The	O
building	O
is	O
now	O
part	O
of	O
the	O
Beth	B-organisation
Israel	I-organisation
Deaconess	I-organisation
Medical	I-organisation
Center	I-organisation
.	O

A	O
common	O
theme	O
of	O
this	O
work	O
is	O
the	O
adoption	O
of	O
a	O
sign-theoretic	O
perspective	O
on	O
issues	O
of	O
artificial	B-field
intelligence	I-field
and	O
knowledge	B-task
representation	I-task
.	O

For	O
instance	O
,	O
the	O
term	O
neural	B-task
machine	I-task
translation	I-task
(	O
NMT	B-task
)	O
emphasizes	O
the	O
fact	O
that	O
deep	O
learning-based	O
approaches	O
to	O
machine	B-task
translation	I-task
directly	O
learn	O
sequence-to-sequence	O
transformations	O
,	O
obviating	O
the	O
need	O
for	O
intermediate	O
steps	O
such	O
as	O
word	B-task
alignment	I-task
and	O
language	B-task
modeling	I-task
that	O
was	O
used	O
in	O
statistical	B-task
machine	I-task
translation	I-task
(	O
SMT	B-task
)	O
.	O

Most	O
research	O
in	O
the	O
field	O
of	O
WSD	B-field
is	O
performed	O
by	O
using	O
WordNet	B-product
as	O
a	O
reference	O
sense	O
inventory	O
for	O
.	O

Notable	O
former	O
PhD	B-misc
students	O
and	O
postdoctoral	O
researchers	O
from	O
his	O
group	O
include	O
Richard	B-researcher
Zemel	I-researcher
,	O
and	O
Zoubin	B-researcher
Ghahramani	I-researcher
.	O

Each	O
prediction	O
result	O
or	O
instance	O
of	O
a	O
confusion	B-metrics
matrix	I-metrics
represents	O
one	O
point	O
in	O
the	O
ROC	B-metrics
space	O
.	O

In	O
1997	O
Thrun	B-researcher
and	O
his	O
colleagues	O
Wolfram	B-researcher
Burgard	I-researcher
and	O
Dieter	B-researcher
Fox	I-researcher
developed	O
the	O
world	O
's	O
first	O
robotic	B-product
tour	I-product
guide	I-product
in	O
the	O
Deutsches	B-location
Museum	I-location
Bonn	I-location
(	O
1997	O
)	O
.	O

WordNet	B-product
is	O
a	O
lexical	O
database	O
of	O
semantic	B-misc
relation	I-misc
s	O
between	O
word	O
s	O
in	O
more	O
than	O
200	O
languages.	O
its	O
primary	O
use	O
is	O
in	O
automatic	O
natural	B-field
language	I-field
processing	I-field
and	O
artificial	B-field
intelligence	I-field
applications	O
.	O

Conferences	O
in	O
the	O
field	O
of	O
natural	B-field
language	I-field
processing	I-field
,	O
such	O
as	O
Association	B-conference
for	I-conference
Computational	I-conference
Linguistics	I-conference
,	O
North	B-conference
American	I-conference
Chapter	I-conference
of	I-conference
the	I-conference
Association	I-conference
for	I-conference
Computational	I-conference
Linguistics	I-conference
,	O
EMNLP	B-conference
,	O
and	O
HLT	B-conference
,	O
are	O
beginning	O
to	O
include	O
papers	O
on	O
speech	B-field
processing	I-field
.	O

A	O
set	O
of	O
Java	B-programlang
programs	O
use	O
the	O
lexicon	O
to	O
work	O
through	O
the	O
variations	O
in	O
biomedical	O
texts	O
by	O
relating	O
words	O
by	O
their	O
parts	B-misc
of	I-misc
speech	I-misc
,	O
which	O
can	O
be	O
helpful	O
in	O
web	O
searches	O
or	O
searches	O
through	O
an	O
electronic	B-misc
medical	I-misc
record	I-misc
.	O

There	O
are	O
many	O
more	O
recent	O
algorithms	O
such	O
as	O
LPBoost	B-algorithm
,	O
TotalBoost	B-algorithm
,	O
BrownBoost	B-algorithm
,	O
xgboost	B-algorithm
,	O
MadaBoost	B-algorithm
,	O
,	O
and	O
others	O
.	O

This	O
is	O
an	O
example	O
implementation	O
in	O
Python	B-programlang
:	O

The	O
Mattel	B-product
Intellivision	B-product
game	O
console	O
offered	O
the	O
Intellivoice	B-task
Voice	I-task
Synthesis	I-task
module	O
in	O
1982	O
.	O

He	O
also	O
worked	O
on	O
machine	B-task
translation	I-task
,	O
both	O
high-accuracy	B-task
knowledge-based	I-task
MT	I-task
and	O
machine	B-field
learning	I-field
for	O
Statistical	B-task
machine	I-task
translation	I-task
(	O
such	O
as	O
generalized	B-task
example-based	I-task
MT	I-task
)	O
.	O

Wolfram	B-organisation
Mathematica	I-organisation
(	O
usually	O
termed	O
Mathematica	B-organisation
)	O
is	O
a	O
modern	O
technical	O
computing	O
system	O
spanning	O
most	O
areas	O
of	O
technical	O
-	O
including	O
neural	B-algorithm
networks	I-algorithm
,	O
machine	B-field
learning	I-field
,	O
image	B-field
processing	I-field
,	O
geometry	B-field
,	O
data	B-field
science	I-field
,	O
visualizations	B-field
,	O
and	O
others	O
.	O

The	O
first	O
digitally	B-product
operated	I-product
and	I-product
programmable	I-product
robot	I-product
was	O
invented	O
by	O
George	B-researcher
Devol	I-researcher
in	O
1954	O
and	O
was	O
ultimately	O
called	O
the	O
Unimate	B-product
.	O

Like	O
DBNs	B-algorithm
,	O
DBMs	B-algorithm
can	O
learn	O
complex	O
and	O
abstract	O
internal	O
representations	O
of	O
the	O
input	O
in	O
tasks	O
such	O
as	O
Object	B-task
recognition	I-task
or	O
speech	B-task
recognition	I-task
,	O
using	O
limited	O
,	O
labeled	O
data	O
to	O
fine-tune	O
the	O
representations	O
built	O
using	O
a	O
large	O
set	O
of	O
unlabeled	O
sensory	O
input	O
data	O
.	O

Scientific	O
conferences	O
where	O
vision	B-task
based	I-task
activity	I-task
recognition	I-task
work	O
often	O
appears	O
are	O
ICCV	B-conference
and	O
CVPR	B-conference
.	O

In	O
statistics	B-field
,	O
an	O
expectation-maximization	B-algorithm
(	O
EM	B-algorithm
)	O
algorithm	O
is	O
an	O
iterative	O
method	O
to	O
find	O
maximum	B-metrics
likelihood	I-metrics
or	O
maximum	B-metrics
a	I-metrics
posteriori	I-metrics
(	O
MAP	B-metrics
)	O
estimates	O
of	O
parameter	O
s	O
in	O
statistical	O
model	O
s	O
,	O
where	O
the	O
model	O
depends	O
on	O
unobserved	O
latent	B-misc
variable	I-misc
s	O
.	O

Similarly	O
,	O
investigators	O
sometimes	O
report	O
the	O
FALSE	B-metrics
Positive	I-metrics
Rate	I-metrics
(	O
FPR	B-metrics
)	O
as	O
well	O
as	O
the	O
FALSE	B-metrics
Negative	I-metrics
Rate	I-metrics
(	O
FNR	B-metrics
)	O
.	O

The	O
concept	O
is	O
similar	O
to	O
the	O
signal	B-metrics
to	I-metrics
noise	I-metrics
ratio	I-metrics
used	O
in	O
the	O
sciences	B-field
and	O
confusion	B-metrics
matrix	I-metrics
used	O
in	O
artificial	B-field
intelligence	I-field
.	O

The	O
Code	O
of	O
Ethics	O
on	O
Human	B-field
Augmentation	I-field
,	O
which	O
was	O
originally	O
introduced	O
by	O
Steve	B-researcher
Mann	I-researcher
in	O
2004	O
and	O
refined	O
with	O
Ray	B-researcher
Kurzweil	I-researcher
and	O
Marvin	B-researcher
Minsky	I-researcher
in	O
2013	O
,	O
was	O
ultimately	O
ratified	O
at	O
the	O
Virtual	B-conference
Reality	I-conference
Toronto	I-conference
conference	I-conference
on	O
June	O
25	O
,	O
2017	O
.	O

In	O
1913	O
,	O
Walter	B-person
R.	I-person
Booth	I-person
directed	O
10	O
films	O
for	O
the	O
U.K.	B-organisation
Kinoplastikon	I-organisation
,	O
presumably	O
in	O
collaboration	O
with	O
Cecil	B-person
Hepworth	I-person
.	O

They	O
introduced	O
their	O
new	O
robot	O
in	O
1961	O
at	O
a	O
trade	O
show	O
at	O
Chicago	B-location
's	O
Cow	B-location
Palace	I-location
.	O

While	O
some	O
chatbot	B-product
applications	O
use	O
extensive	O
word-classification	B-task
processes	O
,	O
natural	B-field
language	I-field
processing	I-field
processors	O
,	O
and	O
sophisticated	O
Artificial	B-field
intelligence	I-field
,	O
others	O
simply	O
scan	O
for	O
general	O
keywords	O
and	O
generate	O
responses	O
using	O
common	O
phrases	O
obtained	O
from	O
an	O
associated	O
library	O
or	O
database	O
.	O

The	O
WaveNet	B-product
model	O
proposed	O
in	O
2016	O
achieves	O
great	O
performance	O
on	O
speech	O
quality	O
.	O

Organizations	O
known	O
to	O
use	O
ALE	B-product
for	O
Emergency	B-misc
management	I-misc
,	O
disaster	B-misc
relief	I-misc
,	O
ordinary	B-misc
communication	I-misc
or	O
extraordinary	B-misc
situation	I-misc
response	I-misc
:	O
American	B-organisation
Red	I-organisation
Cross	I-organisation
,	O
FEMA	B-organisation
,	O
Disaster	B-organisation
Medical	I-organisation
Assistance	I-organisation
Team	I-organisation
s	O
,	O
NATO	B-organisation
,	O
Federal	B-organisation
Bureau	I-organisation
of	I-organisation
Investigation	I-organisation
,	O
United	B-organisation
Nations	I-organisation
,	O
AT	B-organisation
&	I-organisation
T	I-organisation
,	O
Civil	B-organisation
Air	I-organisation
Patrol	I-organisation
,	O
(	O
ARES	B-organisation
)	O
.	O

Here	O
,	O
the	O
Kronecker	B-algorithm
delta	I-algorithm
is	O
used	O
for	O
simplicity	O
(	O
cf.	O
the	O
derivative	O
of	O
a	O
sigmoid	B-algorithm
function	I-algorithm
,	O
being	O
expressed	O
via	O
the	O
function	O
itself	O
)	O
.	O

The	O
theory	O
is	O
based	O
in	O
philosophical	O
foundations	O
,	O
and	O
was	O
founded	O
by	O
Ray	B-researcher
Solomonoff	I-researcher
around	O
1960	O
.	O
Samuel	B-researcher
Rathmanner	I-researcher
and	O
Marcus	B-researcher
Hutter	I-researcher
.	O

WordNet	B-product
,	O
a	O
freely	O
available	O
database	O
originally	O
designed	O
as	O
a	O
semantic	B-misc
network	I-misc
based	O
on	O
psycholinguistic	B-misc
principles	I-misc
,	O
was	O
expanded	O
by	O
addition	O
of	O
definitions	O
and	O
is	O
now	O
also	O
viewed	O
as	O
a	O
dictionary	O
.	O

Advances	O
in	O
the	O
field	O
of	O
computational	B-field
imaging	I-field
research	O
is	O
presented	O
in	O
several	O
venues	O
including	O
publications	O
of	O
SIGGRAPH	B-conference
and	O
the	O
.	O

Classification	B-task
can	O
be	O
thought	O
of	O
as	O
two	O
separate	O
problems	O
-	O
binary	B-task
classification	I-task
and	O
multiclass	B-task
classification	I-task
.	O

Advanced	O
gene	O
finders	O
for	O
both	O
prokaryotic	O
and	O
eukaryotic	O
genomes	O
typically	O
use	O
complex	O
probabilistic	B-algorithm
model	I-algorithm
s	O
,	O
such	O
as	O
hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
s	O
(	O
HMMs	B-algorithm
)	O
to	O
combine	O
information	O
from	O
a	O
variety	O
of	O
different	O
signal	O
and	O
content	O
measurements	O
.	O

Neuroevolution	B-misc
,	O
or	O
neuro-evolution	B-misc
,	O
is	O
a	O
form	O
of	O
artificial	B-field
intelligence	I-field
that	O
uses	O
evolutionary	B-algorithm
algorithm	I-algorithm
s	O
to	O
generate	O
artificial	B-algorithm
neural	I-algorithm
network	I-algorithm
s	O
(	O
ANN	B-algorithm
)	O
,	O
parameters	O
,	O
topology	O
and	O
rules.	O
and	O
evolutionary	B-algorithm
robotics	I-algorithm
.	O

Since	O
IBM	B-organisation
proposed	O
and	O
realized	O
the	O
system	O
of	O
BLEU	B-metrics
Papineni	B-researcher
et	O
al	O
.	O

In	O
2009	O
,	O
experts	O
attended	O
a	O
conference	O
hosted	O
by	O
the	O
Association	B-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Artificial	I-conference
Intelligence	I-conference
(	O
AAAI	B-conference
)	O
to	O
discuss	O
whether	O
computers	O
and	O
robots	O
might	O
be	O
able	O
to	O
acquire	O
any	O
autonomy	O
,	O
and	O
how	O
much	O
these	O
abilities	O
might	O
pose	O
a	O
threat	O
or	O
hazard	O
.	O

After	O
boosting	O
,	O
a	O
classifier	O
constructed	O
from	O
200	O
features	O
could	O
yield	O
a	O
95	O
%	O
detection	O
rate	O
under	O
a	O
^	O
{	O
-5	O
}	O
/	O
math	O
FALSE	B-metrics
positive	I-metrics
rate	I-metrics
.P.	B-researcher
Viola	I-researcher
,	O
M.	B-researcher
Jones	I-researcher
,	O
Robust	B-task
Real-time	I-task
Object	I-task
Detection	I-task
,	O
2001	O
.	O

The	O
website	O
was	O
originally	O
Perl	B-programlang
-based	O
,	O
but	O
IMDb	B-organisation
no	O
longer	O
discloses	O
what	O
software	O
it	O
uses	O
for	O
reasons	O
of	O
security	O
.	O

The	O
start-up	O
was	O
founded	O
by	O
Demis	B-researcher
Hassabis	I-researcher
,	O
Shane	B-researcher
Legg	I-researcher
and	O
Mustafa	B-person
Suleyman	I-person
in	O
2010	O
.	O

Two	O
very	O
commonly	O
used	O
loss	B-misc
functions	I-misc
are	O
the	O
mean	B-metrics
squared	I-metrics
error	I-metrics
,	O
mathL	O
(	O
a	O
)	O
=	O
a	O
^	O
2	O
/	O
math	O
,	O
and	O
the	O
absolute	B-metrics
loss	I-metrics
,	O
mathL	O
(	O
a	O
)	O
=	O
|	O
a	O
|	O
/	O
math	O
.	O

The	O
soft-margin	O
support	B-algorithm
vector	I-algorithm
machine	I-algorithm
described	O
above	O
is	O
an	O
example	O
of	O
an	O
empirical	B-algorithm
risk	I-algorithm
minimization	I-algorithm
(	O
ERM	B-algorithm
)	O
for	O
the	O
hinge	B-metrics
loss	I-metrics
.	O

A	O
deep	B-field
learning	I-field
based	O
approach	O
to	O
MT	B-task
,	O
neural	B-task
machine	I-task
translation	I-task
has	O
made	O
rapid	O
progress	O
in	O
recent	O
years	O
,	O
and	O
Google	B-organisation
has	O
announced	O
its	O
translation	O
services	O
are	O
now	O
using	O
this	O
technology	O
in	O
preference	O
to	O
its	O
previous	O
statistical	O
methods	O
.	O

This	O
tends	O
to	O
yield	O
very	O
large	O
performance	O
gains	O
when	O
working	O
with	O
large	O
corpora	O
such	O
as	O
WordNet	B-product
.	O

Face	B-task
detection	I-task
is	O
used	O
in	O
biometrics	B-field
,	O
often	O
as	O
a	O
part	O
of	O
(	O
or	O
together	O
with	O
)	O
a	O
facial	B-product
recognition	I-product
system	I-product
.	O

trained	O
by	O
maximum	B-algorithm
likelihood	I-algorithm
estimation	I-algorithm
.	O

,	O
Ltd.	O
in	O
Thailand	B-country
;	O
Komatsu	B-organisation
(	I-organisation
Shanghai	I-organisation
)	I-organisation
Ltd.	I-organisation
in	O
1996	O
in	O
Shanghai	B-location
,	O
China	B-country
;	O
Industrial	B-organisation
Power	I-organisation
Alliance	I-organisation
Ltd.	I-organisation
in	O
Japan	B-country
,	O
a	O
joint	O
venture	O
with	O
Cummins	B-organisation
,	O
in	O
1998	O
;	O
L	B-organisation
&	I-organisation
T-Komatsu	I-organisation
Limited	I-organisation
in	O
India	B-country
in	O
1998	O
(	O
shares	O
sold	O
in	O
2013	O
)	O
;	O
and	O
Komatsu	B-organisation
Brasil	I-organisation
International	I-organisation
Ltda.	I-organisation
in	O
Brazil	B-country
in	O
1998	O
.	O

dgp	B-organisation
also	O
occasionally	O
hosts	O
artists	B-misc
in	I-misc
residence	I-misc
(	O
e.g.	O
,	O
Oscar	B-misc
-winner	O
Chris	B-person
Landreth	I-person
.	O

It	O
currently	O
includes	O
four	O
sub-competitions	O
-	O
the	O
RoboMaster	B-misc
Robotics	I-misc
Competition	I-misc
,	O
the	O
RoboMaster	B-misc
Technical	I-misc
Challenge	I-misc
,	O
the	O
ICRA	B-misc
RoboMaster	I-misc
AI	I-misc
Challenge	I-misc
,	O
and	O
the	O
new	O
RoboMaster	B-misc
Youth	I-misc
Tournament	I-misc
.	O

By	O
the	O
early	O
2000s	O
,	O
the	O
dominant	O
speech	B-field
processing	I-field
strategy	O
started	O
to	O
shift	O
away	O
from	O
Hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
towards	O
more	O
modern	O
neural	B-algorithm
networks	I-algorithm
and	O
deep	B-field
learning	I-field
.	O

Another	O
equivalent	O
expression	O
,	O
in	O
the	O
case	O
of	O
a	O
binary	B-metrics
target	I-metrics
rate	I-metrics
,	O
is	O
that	O
the	O
TRUE	B-metrics
positive	I-metrics
rate	I-metrics
and	O
the	O
FALSE	B-metrics
positive	I-metrics
rate	I-metrics
are	O
equal	O
(	O
and	O
therefore	O
the	O
FALSE	B-metrics
negative	I-metrics
rate	I-metrics
and	O
the	O
TRUE	B-metrics
negative	I-metrics
rate	I-metrics
are	O
equal	O
)	O
for	O
every	O
value	O
of	O
the	O
sensitive	O
characteristics	O
:	O

The	O
MATLAB	B-product
function	O
,	O

An	O
articulated	B-product
robot	I-product
is	O
a	O
robot	O
with	O
rotary	B-misc
joint	I-misc
s	O
(	O
e.g.	O
a	O
legged	O
robot	O
or	O
an	O
industrial	B-product
robot	I-product
)	O
.	O

Pandora	B-product
(	O
also	O
known	O
as	O
Pandora	B-product
Media	I-product
or	O
Pandora	B-product
Radio	I-product
)	O
is	O
an	O
American	B-misc
music	O
streaming	O
and	O
automated	B-product
Recommender	I-product
system	I-product
internet	O
radio	O
service	O
powered	O
by	O
the	O
Music	B-misc
Genome	I-misc
Project	I-misc
and	O
headquartered	O
in	O
Oakland	B-location
,	O
California	B-location
.	O

She	O
is	O
a	O
board	O
member	O
of	O
the	O
International	B-organisation
Machine	I-organisation
Learning	I-organisation
Society	I-organisation
,	O
has	O
been	O
a	O
member	O
of	O
AAAI	B-organisation
Executive	I-organisation
council	I-organisation
,	O
was	O
PC	O
co-chair	O
of	O
ICML	B-conference
2011	I-conference
,	O
and	O
has	O
served	O
as	O
senior	O
PC	O
member	O
for	O
conferences	O
including	O
AAAI	B-conference
,	O
ICML	B-conference
,	O
IJCAI	B-conference
,	O
ISWC	B-conference
,	O
KDD	B-conference
,	O
SIGMOD	B-conference
,	O
UAI	B-conference
,	O
VLDB	B-conference
,	O
WSDM	B-conference
and	O
WWW	B-conference
.	O

James	B-researcher
S.	I-researcher
Albus	I-researcher
of	O
the	O
National	B-organisation
Institute	I-organisation
of	I-organisation
Standards	I-organisation
and	I-organisation
Technology	I-organisation
(	O
NIST	B-organisation
)	O
developed	O
the	O
Robocrane	B-product
,	O
where	O
the	O
platform	O
hangs	O
from	O
six	O
cables	O
instead	O
of	O
being	O
supported	O
by	O
six	O
jacks	O
.	O

Another	O
class	O
of	O
direct	B-misc
search	I-misc
algorithms	I-misc
are	O
the	O
various	O
evolutionary	B-algorithm
algorithm	I-algorithm
s	O
,	O
e.g.	O
genetic	B-algorithm
algorithm	I-algorithm
s	O
.	O

KUKA	B-organisation
is	O
a	O
German	B-misc
manufacturer	O
of	O
industrial	B-product
robot	I-product
s	O
and	O
solution	O
s	O
for	O
factory	O
automation	O
.	O

Other	O
films	O
between	O
2016	O
to	O
2020	O
that	O
captured	O
with	O
IMAX	B-misc
camera	O
's	O
were	O
Zack	B-person
Snyder	I-person
'	O
s	O
Batman	B-misc
v	I-misc
Superman	I-misc
:	I-misc
Dawn	I-misc
of	I-misc
Justice	I-misc
,	O
Clint	B-person
Eastwood	I-person
'	O
s	O
Sully	B-misc
,	O
Damien	B-person
Chazelle	I-person
'	O
s	O
First	B-misc
Man	I-misc
,	O
Patty	B-person
Jenkins	I-person
'	O
Wonder	B-misc
Woman	I-misc
1984	I-misc
,	O
Cary	B-person
Joji	I-person
Fukunaga	I-person
'	O
s	O
No	B-misc
Time	I-misc
to	I-misc
Die	I-misc
and	O
Joseph	B-person
Kosinski	I-person
'	O
s	O
Top	B-misc
Gun	I-misc
:	I-misc
Maverick	I-misc
.	O

The	O
trial	O
of	O
MICR	B-misc
E13B	I-misc
font	O
was	O
shown	O
to	O
the	O
American	B-organisation
Bankers	I-organisation
Association	I-organisation
(	O
ABA	B-organisation
)	O
in	O
July	O
1956	O
,	O
which	O
adopted	O
it	O
in	O
1958	O
as	O
the	O
MICR	B-misc
standard	O
for	O
negotiable	O
document	O
s	O
in	O
the	O
United	B-country
States	I-country
.	O

Local	B-misc
search	I-misc
algorithms	I-misc
are	O
widely	O
applied	O
to	O
numerous	O
hard	O
computational	O
problems	O
,	O
including	O
problems	O
from	O
computer	B-field
science	I-field
(	O
particularly	O
artificial	B-field
intelligence	I-field
)	O
,	O
mathematics	B-field
,	O
operations	B-field
research	I-field
,	O
engineering	B-field
,	O
and	O
bioinformatics	B-field
.	O

Gerd	B-researcher
Gigerenzer	I-researcher
(	O
born	O
September	O
3	O
,	O
1947	O
,	O
Wallersdorf	B-location
,	O
Germany	B-country
)	O
is	O
a	O
Germany	B-country
psychologist	O
who	O
has	O
studied	O
the	O
use	O
of	O
bounded	B-algorithm
rationality	I-algorithm
and	O
heuristic	B-algorithm
s	O
in	O
decision	B-task
making	I-task
.	O

to	O
minimize	O
the	O
Mean	B-metrics
squared	I-metrics
error	I-metrics
.	O

But	O
even	O
an	O
official	O
language	O
with	O
a	O
regulating	O
academy	O
,	O
such	O
as	O
Standard	B-misc
French	I-misc
with	O
the	O
Académie	B-organisation
française	I-organisation
,	O
is	O
classified	O
as	O
a	O
natural	O
language	O
(	O
for	O
example	O
,	O
in	O
the	O
field	O
of	O
natural	B-field
language	I-field
processing	I-field
)	O
,	O
as	O
its	O
prescriptive	O
points	O
do	O
not	O
make	O
it	O
either	O
constructed	O
enough	O
to	O
be	O
classified	O
as	O
a	O
constructed	B-misc
language	I-misc
or	O
controlled	O
enough	O
to	O
be	O
classified	O
as	O
a	O
controlled	B-misc
natural	I-misc
language	I-misc
.	O

There	O
are	O
a	O
number	O
of	O
other	O
metrics	O
,	O
most	O
simply	O
the	O
accuracy	B-metrics
or	O
Fraction	B-metrics
Correct	I-metrics
(	O
FC	B-metrics
)	O
,	O
which	O
measures	O
the	O
fraction	O
of	O
all	O
instances	O
that	O
are	O
correctly	O
categorized	O
;	O
the	O
complement	O
is	O
the	O
Fraction	B-metrics
Incorrect	I-metrics
(	O
FiC	B-metrics
)	O
.	O

Cardie	B-researcher
became	O
a	O
Fellow	O
of	O
the	O
Association	B-conference
for	I-conference
Computational	I-conference
Linguistics	I-conference
in	O
2016	O
.	O

Learning	O
the	O
parameters	O
math	O
\	O
theta	O
/	O
math	O
is	O
usually	O
done	O
by	O
maximum	B-algorithm
likelihood	I-algorithm
learning	I-algorithm
for	O
mathp	O
(	O
Y	O
_	O
i	O
|	O
X	O
_	O
i	O
;	O
\	O
theta	O
)	O
/	O
math	O
.	O

Cluster	B-task
analysis	I-task
,	O
and	O
Non-negative	B-algorithm
matrix	I-algorithm
factorization	I-algorithm
for	O
descriptive	B-task
mining	I-task
.	O

In	O
computer	B-field
science	I-field
and	O
the	O
information	B-field
technology	I-field
that	O
it	O
enables	O
,	O
it	O
has	O
been	O
a	O
long-term	O
challenge	O
to	O
the	O
ability	O
in	O
computers	O
to	O
do	O
natural	B-field
language	I-field
processing	I-field
and	O
machine	B-field
learning	I-field
.	O

(	O
Code	O
for	O
Gabor	B-algorithm
feature	I-algorithm
extraction	I-algorithm
from	O
images	O
in	O
MATLAB	B-product
can	O
be	O
found	O
at	O

The	O
NeuralExpert	B-misc
centers	O
the	O
design	O
specifications	O
around	O
the	O
type	O
of	O
problem	O
the	O
user	O
would	O
like	O
the	O
neural	B-algorithm
network	I-algorithm
to	O
solve	O
(	O
Classification	B-task
,	O
Prediction	B-task
,	O
Function	B-task
approximation	I-task
or	O
Cluster	B-task
analysis	I-task
)	O
.	O

When	O
the	O
quantization	B-misc
step	I-misc
size	I-misc
(	O
Δ	O
)	O
is	O
small	O
relative	O
to	O
the	O
variation	O
in	O
the	O
signal	O
being	O
quantized	O
,	O
it	O
is	O
relatively	O
simple	O
to	O
show	O
that	O
the	O
mean	B-metrics
squared	I-metrics
error	I-metrics
produced	O
by	O
such	O
a	O
rounding	O
operation	O
will	O
be	O
approximately	O
math	O
\	O
Delta	O
^	O
2	O
/	O
12	O
/	O
math.math	O

The	O
construction	O
of	O
a	O
rich	O
lexicon	O
with	O
a	O
suitable	O
ontology	O
requires	O
significant	O
effort	O
,	O
e.g.	O
,	O
Wordnet	B-product
lexicon	O
required	O
many	O
person-years	O
of	O
effort.	O
G.	B-researcher
A.	I-researcher
Miller	I-researcher
,	O
R.	B-researcher
Beckwith	I-researcher
,	O
C.	B-researcher
D.	I-researcher
Fellbaum	I-researcher
,	O
D.	B-researcher
Gross	I-researcher
,	O
K.	B-researcher
Miller	I-researcher
.	O

Kawasaki	B-organisation
's	O
portfolio	O
also	O
includes	O
retractable	O
roofs	O
,	O
floors	O
and	O
other	O
giant	O
structures	O
,	O
the	O
Sapporo	B-location
Dome	I-location
'	O
retractable	O
surface	O
is	O
one	O
example	O
.	O

Kappa	B-metrics
statistics	I-metrics
such	O
as	O
Fleiss	B-metrics
'	I-metrics
kappa	I-metrics
and	O
Cohen	B-metrics
's	I-metrics
kappa	I-metrics
are	O
methods	O
for	O
calculating	O
inter-rater	B-metrics
reliability	I-metrics
based	O
on	O
different	O
assumptions	O
about	O
the	O
marginal	O
or	O
prior	O
distributions	O
,	O
and	O
are	O
increasingly	O
used	O
as	O
chance	O
corrected	O
alternatives	O
to	O
accuracy	B-metrics
in	O
other	O
contexts	O
.	O

With	O
his	O
students	O
Sepp	B-researcher
Hochreiter	I-researcher
,	O
Felix	B-researcher
Gers	I-researcher
,	O
Fred	B-researcher
Cummins	I-researcher
,	O
Alex	B-researcher
Graves	I-researcher
,	O
and	O
others	O
,	O
Schmidhuber	B-researcher
published	O
increasingly	O
sophisticated	O
versions	O
of	O
a	O
type	O
of	O
recurrent	B-algorithm
neural	I-algorithm
network	I-algorithm
called	O
the	O
long	B-algorithm
short-term	I-algorithm
memory	I-algorithm
(	O
LSTM	B-algorithm
)	O
.	O

2004	O
-	O
The	O
first	O
Cobot	B-product
KUKA	I-product
LBR	I-product
3	I-product
is	O
released	O
.	O

Two	O
shallow	O
approaches	O
used	O
to	O
train	O
and	O
then	O
disambiguate	O
are	O
Naive	B-algorithm
Bayes	I-algorithm
classifier	I-algorithm
and	O
decision	B-algorithm
trees	I-algorithm
.	O

The	O
first	O
practical	O
forms	O
of	O
photography	B-misc
were	O
introduced	O
in	O
January	O
1839	O
by	O
Louis	B-person
Daguerre	I-person
and	O
Henry	B-person
Fox	I-person
Talbot	I-person
.	O

For	O
example	O
,	O
speech	B-task
synthesis	I-task
,	O
combined	O
with	O
speech	B-task
recognition	I-task
,	O
allows	O
for	O
interaction	O
with	O
mobile	O
devices	O
via	O
language	B-field
processing	I-field
interfaces	O
.	O

Phidgets	B-product
can	O
be	O
programmed	O
using	O
a	O
variety	O
of	O
software	O
and	O
programming	O
languages	O
,	O
ranging	O
from	O
Java	B-programlang
to	O
Microsoft	B-product
Excel	I-product
.	O

The	O
term	O
machine	B-field
learning	I-field
was	O
coined	O
in	O
1959	O
by	O
Arthur	B-researcher
Samuel	I-researcher
,	O
an	O
American	B-misc
IBMer	I-misc
and	O
pioneer	O
in	O
the	O
field	O
of	O
computer	B-field
gaming	I-field
and	O
artificial	B-field
intelligence	I-field
.	O

The	O
Israeli	B-misc
poet	O
David	B-person
Avidan	I-person
,	O
who	O
was	O
fascinated	O
with	O
future	O
technologies	O
and	O
their	O
relation	O
to	O
art	O
,	O
desired	O
to	O
explore	O
the	O
use	O
of	O
computers	O
for	O
writing	O
literature	O
.	O

As	O
part	O
of	O
the	O
GATEway	B-misc
Project	I-misc
in	O
2017	O
,	O
Oxbotica	B-organisation
trialled	O
seven	O
autonomous	O
shuttle	O
buses	O
in	O
Greenwich	B-location
,	O
navigating	O
a	O
two-mile	O
riverside	O
path	O
near	O
London	B-location
's	O
The	B-location
O2	I-location
Arena	I-location
on	O
a	O
route	O
also	O
used	O
by	O
pedestrians	O
and	O
cyclists	O
.	O

An	O
unrelated	O
but	O
commonly	O
used	O
combination	O
of	O
basic	O
statistics	O
from	O
information	B-task
retrieval	I-task
is	O
the	O
F-score	B-metrics
,	O
being	O
a	O
(	O
possibly	O
weighted	O
)	O
harmonic	B-misc
mean	I-misc
of	O
recall	B-metrics
and	O
precision	B-metrics
where	O
recall	B-metrics
=	O
sensitivity	B-metrics
=	O
TRUE	B-metrics
positive	I-metrics
rate	I-metrics
,	O
but	O
specificity	B-metrics
and	O
precision	B-metrics
are	O
totally	O
different	O
measures	O
.	O

Neuromorphic	B-field
engineering	I-field
is	O
an	O
interdisciplinary	O
subject	O
that	O
takes	O
inspiration	O
from	O
biology	B-field
,	O
physics	B-field
,	O
mathematics	B-field
,	O
computer	B-field
science	I-field
,	O
and	O
electronic	B-field
engineering	I-field
to	O
design	O
artificial	O
neural	O
systems	O
,	O
such	O
as	O
vision	B-product
systems	I-product
,	O
head-eye	B-product
systems	I-product
,	O
auditory	B-product
processors	I-product
,	O
and	O
autonomous	B-product
robots	I-product
,	O
whose	O
physical	O
architecture	O
and	O
design	O
principles	O
are	O
based	O
on	O
those	O
of	O
biological	B-product
nervous	I-product
systems	I-product
.	O

To	O
be	O
specific	O
,	O
the	O
BIBO	B-metrics
stability	I-metrics
criterion	I-metrics
requires	O
that	O
the	O
ROC	B-metrics
of	O
the	O
system	O
includes	O
the	O
unit	O
circle	O
.	O

2	O
The	O
program	O
was	O
rewritten	O
in	O
Java	B-programlang
beginning	O
in	O
1998	O
.	O

The	O
MCC	B-metrics
can	O
be	O
calculated	O
directly	O
from	O
the	O
confusion	B-metrics
matrix	I-metrics
using	O
the	O
formula	O
:	O

It	O
was	O
developed	O
by	O
a	O
team	O
at	O
the	O
MIT-IBM	B-organisation
Watson	I-organisation
AI	I-organisation
Lab	I-organisation
and	O
first	O
presented	O
at	O
the	O
2018	B-conference
International	I-conference
Conference	I-conference
on	I-conference
Learning	I-conference
Representations	I-conference
.	O

When	O
the	O
TRUE	B-metrics
prevalence	I-metrics
s	O
for	O
the	O
two	O
positive	O
variables	O
are	O
equal	O
as	O
assumed	O
in	O
Fleiss	B-metrics
kappa	I-metrics
and	O
F-score	B-metrics
,	O
that	O
is	O
the	O
number	O
of	O
positive	O
predictions	O
matches	O
the	O
number	O
of	O
positive	O
classes	O
in	O
the	O
dichotomous	O
(	O
two	O
class	O
)	O
case	O
,	O
the	O
different	O
kappa	B-metrics
and	O
correlation	B-metrics
measure	O
collapse	O
to	O
identity	O
with	O
Youden	B-researcher
's	I-researcher
J	I-researcher
,	O
and	O
recall	B-metrics
,	O
precision	B-metrics
and	O
F-score	B-metrics
are	O
similarly	O
identical	O
with	O
accuracy	B-metrics
.	O

The	O
Building	B-conference
Educational	I-conference
Applications	I-conference
workshop	I-conference
(	O
BEA	B-conference
)	O
at	O
NAACL	B-conference
2013	O
hosted	O
the	O
inaugural	O
NLI	B-task
shared	I-task
task.	I-task
Tetreault	B-researcher
et	O
al	O
,	O
2013	O
The	O
competition	O
resulted	O
in	O
29	O
entries	O
from	O
teams	O
across	O
the	O
globe	O
,	O
24	O
of	O
which	O
also	O
published	O
a	O
paper	O
describing	O
their	O
systems	O
and	O
approaches	O
.	O

The	O
Viterbi	B-algorithm
algorithm	I-algorithm
is	O
a	O
dynamic	B-algorithm
programming	I-algorithm
algorithm	I-algorithm
for	O
finding	O
the	O
most	O
likely	O
sequence	O
of	O
hidden	B-misc
states	I-misc
called	O
the	O
Viterbi	B-misc
path	I-misc
that	O
results	O
in	O
a	O
sequence	O
of	O
observed	O
events	O
,	O
especially	O
in	O
the	O
context	O
of	O
Markov	B-misc
information	I-misc
source	I-misc
s	O
and	O
hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
s	O
(	O
HMM	B-algorithm
)	O
.	O

In	O
statistics	B-field
,	O
multinomial	B-algorithm
logistic	I-algorithm
regression	I-algorithm
is	O
a	O
classification	B-misc
method	I-misc
that	O
generalizes	O
logistic	B-algorithm
regression	I-algorithm
to	O
multiclass	B-task
classification	I-task
,	O
i.e.	O
with	O
more	O
than	O
two	O
possible	O
discrete	O
outcomes	O
.	O

Hidden	B-algorithm
Markov	I-algorithm
models	I-algorithm
are	O
known	O
for	O
their	O
applications	O
to	O
reinforcement	B-field
learning	I-field
and	O
temporal	B-field
pattern	I-field
recognition	I-field
such	O
as	O
speech	B-task
,	O
handwriting	B-task
recognition	I-task
,	O
gesture	B-task
recognition	I-task
,	O
Thad	B-researcher
Starner	I-researcher
,	O
Alex	B-researcher
Pentland	I-researcher
.	O

Essentially	O
,	O
this	O
means	O
that	O
if	O
the	O
n-gram	B-misc
has	O
been	O
seen	O
more	O
than	O
k	O
times	O
in	O
training	O
,	O
the	O
conditional	O
probability	O
of	O
a	O
word	O
given	O
its	O
history	O
is	O
proportional	O
to	O
the	O
maximum	B-metrics
likelihood	I-metrics
estimate	I-metrics
of	O
that	O
n	B-misc
-gram	I-misc
.	O

He	O
is	O
interested	O
in	O
knowledge	B-task
representation	I-task
,	O
commonsense	B-task
reasoning	I-task
,	O
and	O
natural	B-task
language	I-task
understanding	I-task
,	O
believing	O
that	O
deep	B-task
language	I-task
understanding	I-task
can	O
only	O
currently	O
be	O
achieved	O
by	O
significant	O
hand-engineering	B-misc
of	O
semantically-rich	O
formalisms	O
coupled	O
with	O
statistical	O
preferences	O
.	O

In	O
JavaScript	B-programlang
,	O
Python	B-programlang
or	O

The	O
Newcomb	B-misc
Awards	I-misc
are	O
announced	O
in	O
the	O
AI	B-misc
Magazine	I-misc
published	O
by	O
AAAI	B-conference
.	O

The	O
Mean	B-metrics
squared	I-metrics
error	I-metrics
on	O
a	O
test	O
set	O
of	O
100	O
exemplars	O
is	O
0.084	O
,	O
smaller	O
than	O
the	O
unnormalized	O
error	O
.	O

The	O
F-score	B-metrics
has	O
been	O
widely	O
used	O
in	O
the	O
natural	B-field
language	I-field
processing	I-field
literature	O
,	O
such	O
as	O
the	O
evaluation	O
of	O
named	B-task
entity	I-task
recognition	I-task
(	O
NER	B-task
)	O
and	O
word	B-task
segmentation	I-task
.	O

Chatbots	B-product
are	O
typically	O
used	O
in	O
dialog	B-product
systems	I-product
for	O
various	O
purposes	O
including	O
customer	O
service	O
,	O
request	B-misc
routing	I-misc
,	O
or	O
for	O
information	B-misc
gathering	I-misc
.	O

Important	O
journals	O
include	O
the	O
IEEE	B-conference
Transactions	I-conference
on	I-conference
Speech	I-conference
and	I-conference
Audio	I-conference
Processing	I-conference
(	O
later	O
renamed	O
IEEE	B-conference
Transactions	I-conference
on	I-conference
Audio	I-conference
,	I-conference
Speech	I-conference
and	I-conference
Language	I-conference
Processing	I-conference
and	O
since	O
Sept	O
2014	O
renamed	O
IEEE	B-conference
/	I-conference
ACM	I-conference
Transactions	I-conference
on	I-conference
Audio	I-conference
,	I-conference
Speech	I-conference
and	I-conference
Language	I-conference
Processing	I-conference
-	O
after	O
merging	O
with	O
an	O
ACM	B-conference
publication	O
)	O
,	O
Computer	B-conference
Speech	I-conference
and	I-conference
Language	I-conference
,	O
and	O
Speech	B-conference
Communication	I-conference
.	O

EM	B-algorithm
is	O
frequently	O
used	O
for	O
data	B-task
clustering	I-task
in	O
machine	B-field
learning	I-field
and	O
computer	B-field
vision	I-field
.	O

While	O
there	O
is	O
no	O
perfect	O
way	O
of	O
describing	O
the	O
confusion	B-metrics
matrix	I-metrics
of	O
TRUE	O
and	O
FALSE	O
positives	O
and	O
negatives	O
by	O
a	O
single	O
number	O
,	O
the	O
Matthews	B-metrics
correlation	I-metrics
coefficient	I-metrics
is	O
generally	O
regarded	O
as	O
being	O
one	O
of	O
the	O
best	O
such	O
measures	O
.	O

As	O
data	O
set	O
s	O
have	O
grown	O
in	O
size	O
and	O
complexity	O
,	O
direct	O
hands-on	O
data	B-field
analysis	I-field
has	O
been	O
augmented	O
with	O
indirect	O
,	O
automated	O
data	O
processing	O
,	O
aided	O
by	O
other	O
discoveries	O
in	O
computer	B-field
science	I-field
,	O
specially	O
in	O
the	O
field	O
of	O
machine	B-field
learning	I-field
,	O
such	O
as	O
neural	B-algorithm
networks	I-algorithm
,	O
cluster	B-task
analysis	I-task
,	O
genetic	B-algorithm
algorithms	I-algorithm
(	O
1950s	O
)	O
,	O
decision	B-algorithm
tree	I-algorithm
learning	I-algorithm
and	O
decision	B-algorithm
rules	I-algorithm
(	O
1960s	O
)	O
,	O
and	O
support	B-algorithm
vector	I-algorithm
machines	I-algorithm
(	O
1990s	O
)	O
.	O

In	O
the	O
fall	O
of	O
2005	O
,	O
Thrun	B-researcher
published	O
a	O
textbook	O
entitled	O
Probabilistic	B-misc
Robotics	I-misc
together	O
with	O
his	O
long-term	O
co-workers	O
Dieter	B-researcher
Fox	I-researcher
and	O
Wolfram	B-researcher
Burgard	I-researcher
.	O

John	B-researcher
D.	I-researcher
Lafferty	I-researcher
,	O
Andrew	B-researcher
McCallum	I-researcher
and	O
Pereiramath	B-researcher
as	O
follows	O
:	O

Question	B-task
answering	I-task
(	O
QA	B-task
)	O
is	O
a	O
computer	B-field
science	I-field
discipline	O
within	O
the	O
fields	O
of	O
information	B-field
retrieval	I-field
and	O
natural	B-field
language	I-field
processing	I-field
(	O
NLP	B-field
)	O
,	O
which	O
is	O
concerned	O
with	O
building	O
systems	O
that	O
automatically	O
answer	O
questions	O
posed	O
by	O
humans	O
in	O
a	O
natural	O
language	O
.	O

However	O
,	O
in	O
the	O
version	O
of	O
the	O
metric	O
used	O
by	O
NIST	B-metrics
evaluations	O
prior	O
to	O
2009	O
,	O
the	O
shortest	O
reference	O
sentence	O
had	O
been	O
used	O
instead	O
.	O

On	O
August	O
27	O
,	O
2018	O
,	O
Toyota	B-person
announced	O
an	O
investment	O
of	O
$	O
500	O
Million	O
in	O
Uber	B-organisation
'	O
s	O
autonomous	B-product
car	I-product
s	O
.	O

The	O
sample	O
maximum	O
is	O
the	O
maximum	B-metrics
likelihood	I-metrics
estimator	I-metrics
for	O
the	O
population	O
maximum	O
,	O
but	O
,	O
as	O
discussed	O
above	O
,	O
it	O
is	O
biased	O
.	O

LSI	B-task
helps	O
overcome	O
synonymy	B-misc
by	O
increasing	O
recall	B-metrics
,	O
one	O
of	O
the	O
most	O
problematic	O
constraints	O
of	O
Boolean	B-algorithm
keyword	I-algorithm
queries	I-algorithm
and	O
vector	B-algorithm
space	I-algorithm
models	I-algorithm
.	O

Data	B-task
acquisition	I-task
applications	O
are	O
usually	O
controlled	O
by	O
software	O
programs	O
developed	O
using	O
various	O
general	O
purpose	O
programming	O
languages	O
such	O
as	O
Assembly	B-programlang
,	O
BASIC	B-programlang
,	O
C	B-programlang
,	O
C	B-programlang
+	I-programlang
+	I-programlang
,	O
C	B-programlang
#	I-programlang
,	O
Fortran	B-programlang
,	O
Java	B-programlang
,	O
LabVIEW	B-programlang
,	O
Lisp	B-programlang
,	O
Pascal	B-programlang
,	O
etc	O
.	O

In	O
2003	O
,	O
Honda	B-organisation
released	O
its	O
Cog	B-product
advertisement	O
in	O
the	O
UK	B-country
and	O
on	O
the	O
Internet	O
.	O

The	O
Association	B-conference
for	I-conference
Computational	I-conference
Linguistics	I-conference
defines	O
computational	B-field
linguistics	I-field
as	O
:	O

Expectation-maximization	B-algorithm
algorithm	I-algorithm
s	O
may	O
be	O
employed	O
to	O
calculate	O
approximate	O
maximum	B-algorithm
likelihood	I-algorithm
estimates	I-algorithm
of	O
unknown	O
state-space	O
parameters	O
within	O
minimum-variance	O
filters	O
and	O
smoothers	O
.	O

Correspondents	O
included	O
former	O
Baywatch	B-misc
actresses	O
Donna	B-person
D	I-person
'Errico	I-person
,	O
Carmen	B-person
Electra	I-person
,	O
and	O
Traci	B-person
Bingham	I-person
,	O
former	O
Playboy	B-misc
Playmate	I-misc
Heidi	B-person
Mark	I-person
,	O
comedian	O
Arj	B-person
Barker	I-person
and	O
identical	O
twins	O
Randy	B-person
and	O
Jason	B-person
Sklar	I-person
.	O

It	O
is	O
commonly	O
used	O
to	O
generate	O
representations	O
for	O
speech	B-task
recognition	I-task
(	O
ASR	B-task
)	O
,	O
e.g.	O
the	O
CMU	B-product
Sphinx	I-product
system	I-product
,	O
and	O
speech	B-task
synthesis	I-task
(	O
TTS	B-task
)	O
,	O
e.g.	O
the	O
Festival	B-product
system	I-product
.	O

Sensitivity	B-metrics
or	O
TRUE	B-metrics
Positive	I-metrics
Rate	I-metrics
(	O
TPR	B-metrics
)	O
,	O
also	O
known	O
as	O
recall	B-metrics
,	O
is	O
the	O
proportion	O
of	O
people	O
that	O
tested	O
positive	O
and	O
are	O
positive	O
(	O
TRUE	B-metrics
Positive	I-metrics
,	O
TP	B-metrics
)	O
of	O
all	O
the	O
people	O
that	O
actually	O
are	O
positive	O
(	O
Condition	B-metrics
Positive	I-metrics
,	O
CP	B-metrics
=	O
TP	B-metrics
+	I-metrics
FN	I-metrics
)	O
.	O

Popular	O
speech	B-task
recognition	I-task
conferences	O
held	O
each	O
year	O
or	O
two	O
include	O
SpeechTEK	B-conference
and	O
SpeechTEK	B-conference
Europe	I-conference
,	O
ICASSP	B-conference
,	O
Interspeech	B-conference
/	O
Eurospeech	B-conference
,	O
and	O
the	O
IEEE	B-conference
ASRU	I-conference
.	O

Devol	B-researcher
collaborated	O
with	O
Engelberger	B-researcher
,	O
who	O
served	O
as	O
president	O
of	O
the	O
company	O
,	O
to	O
engineer	O
and	O
produce	O
an	O
industrial	B-product
robot	I-product
under	O
the	O
brand	O
name	O
Unimate	B-product
.	O

A	O
Hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
(	O
HMM	B-algorithm
)	O
is	O
a	O
statistical	B-algorithm
Markov	I-algorithm
model	I-algorithm
in	O
which	O
the	O
system	O
being	O
modeled	O
is	O
assumed	O
to	O
be	O
a	O
Markov	B-algorithm
process	I-algorithm
with	O
unobserved	O
(	O
hidden	O
)	O
states	O
.	O

This	O
property	O
,	O
undesirable	O
in	O
many	O
applications	O
,	O
has	O
led	O
researchers	O
to	O
use	O
alternatives	O
such	O
as	O
the	O
mean	B-metrics
absolute	I-metrics
error	I-metrics
,	O
or	O
those	O
based	O
on	O
the	O
median	B-misc
.	O

Such	O
a	O
sequence	O
(	O
which	O
depends	O
on	O
the	O
outcome	O
of	O
the	O
investigation	O
of	O
previous	O
attributes	O
at	O
each	O
stage	O
)	O
is	O
called	O
a	O
decision	B-algorithm
tree	I-algorithm
and	O
applied	O
in	O
the	O
area	O
of	O
machine	B-field
learning	I-field
known	O
as	O
decision	B-algorithm
tree	I-algorithm
learning	I-algorithm
.	O

As	O
in	O
factor	B-task
analysis	I-task
,	O
the	O
LCA	B-algorithm
can	O
also	O
be	O
used	O
to	O
classify	O
case	O
according	O
to	O
their	O
maximum	B-algorithm
likelihood	I-algorithm
class	O
membership	O
.	O

Supervised	B-algorithm
neural	I-algorithm
networks	I-algorithm
that	O
use	O
a	O
mean	B-metrics
squared	I-metrics
error	I-metrics
(	O
MSE	B-metrics
)	O
cost	B-misc
function	I-misc
can	O
use	O
formal	O
statistical	O
methods	O
to	O
determine	O
the	O
confidence	O
of	O
the	O
trained	O
model	O
.	O

This	O
can	O
be	O
directly	O
expressed	O
as	O
a	O
linear	O
program	O
,	O
but	O
it	O
is	O
also	O
equivalent	O
to	O
Tikhonov	B-algorithm
regularization	I-algorithm
with	O
the	O
hinge	B-metrics
loss	I-metrics
function	I-metrics
,	O
mathV	O
(	O
f	O
(	O
x	O
)	O
,	O
y	O
)	O
=	O
\	O
max	O
(	O
0	O
,	O
1	O
-	O
yf	O
(	O
x	O
)	O
)	O
/	O
math	O
:	O

The	O
following	O
technique	O
was	O
described	O
in	O
Breiman	B-researcher
's	O
original	O
paper	O
and	O
is	O
implemented	O
in	O
the	O
R	B-product
package	I-product
randomForest	I-product
.	O

Traditional	O
image	O
quality	O
measures	O
,	O
such	O
as	O
PSNR	B-metrics
,	O
are	O
typically	O
performed	O
on	O
fixed	O
resolution	O
images	O
and	O
do	O
not	O
take	O
into	O
account	O
some	O
aspects	O
of	O
the	O
human	O
visual	O
system	O
,	O
like	O
the	O
change	O
in	O
spatial	O
resolution	O
across	O
the	O
retina	B-misc
.	O

John	B-person
Ireland	I-person
,	O
Joanne	B-person
Dru	I-person
and	O
Macdonald	B-person
Carey	I-person
starred	O
in	O
the	O
Jack	B-person
Broder	I-person
color	O
production	O
Hannah	B-misc
Lee	I-misc
,	O
which	O
premiered	O
June	O
19	O
,	O
1953	O
.	O

That	O
process	O
is	O
called	O
image	B-task
registration	I-task
,	O
and	O
uses	O
different	O
methods	O
of	O
computer	B-field
vision	I-field
,	O
mostly	O
related	O
to	O
tracking	B-task
.	O

Now	O
let	O
us	O
start	O
explaining	O
the	O
different	O
possible	O
relations	O
between	O
predicted	O
and	O
actual	O
outcome	O
:	O
Confusion	B-metrics
matrix	I-metrics

The	O
VOICEBOX	B-product
speech	B-misc
processing	I-misc
toolbox	I-misc
for	O
MATLAB	B-product
implements	O
the	O
conversion	O
and	O
its	O
inverse	O
as	O
:	O

Prolog	B-programlang
is	O
a	O
logic	O
programming	O
language	O
associated	O
with	O
artificial	B-field
intelligence	I-field
and	O
computational	B-field
linguistics	I-field
.	O

Milner	B-researcher
has	O
received	O
numerous	O
awards	O
for	O
her	O
contributions	O
to	O
neuroscience	B-field
and	O
psychology	B-field
including	O
memberships	O
in	O
the	O
Royal	B-organisation
Society	I-organisation
of	I-organisation
London	I-organisation
,	O
the	O
Royal	B-organisation
Society	I-organisation
of	I-organisation
Canada	I-organisation
and	O
the	O
National	B-organisation
Academy	I-organisation
of	I-organisation
Sciences	I-organisation
.	O

By	O
combining	O
these	O
operators	O
one	O
can	O
obtain	O
algorithms	O
for	O
many	O
image	B-field
processing	I-field
tasks	O
,	O
such	O
as	O
feature	B-task
extraction	I-task
,	O
image	B-task
segmentation	I-task
,	O
image	B-task
sharpening	I-task
,	O
image	B-task
filtering	I-task
,	O
and	O
classification	B-task
.	O

As	O
of	O
2017	O
,	O
he	O
is	O
a	O
professor	O
at	O
the	O
Collège	B-university
de	I-university
France	I-university
and	O
,	O
since	O
1989	O
,	O
the	O
director	O
of	O
INSERM	B-organisation
Unit	I-organisation
562	I-organisation
,	O
Cognitive	B-field
Neuroimaging	I-field
.	O

There	O
are	O
many	O
approaches	O
to	O
learning	O
these	O
embeddings	O
,	O
notably	O
using	O
Bayesian	B-algorithm
clustering	I-algorithm
frameworks	I-algorithm
or	O
energy-based	B-algorithm
frameworks	I-algorithm
,	O
and	O
more	O
recently	O
,	O
TransE	B-conference
(	O
Conference	B-conference
on	I-conference
Neural	I-conference
Information	I-conference
Processing	I-conference
Systems	I-conference
2013	I-conference
)	O
.	O

It	O
is	O
an	O
alternative	O
to	O
the	O
Word	B-metrics
error	I-metrics
rate	I-metrics
(	O
Word	B-metrics
Error	I-metrics
Rate	I-metrics
)	O
used	O
in	O
several	O
countries	O
.	O

ANNs	B-algorithm
have	O
been	O
used	O
on	O
a	O
variety	O
of	O
tasks	O
,	O
including	O
computer	B-field
vision	I-field
,	O
speech	B-task
recognition	I-task
,	O
machine	B-task
translation	I-task
,	O
social	B-task
network	I-task
filtering	I-task
,	O
playing	B-task
board	I-task
and	I-task
video	I-task
games	I-task
,	O
medical	B-task
diagnosis	I-task
,	O
and	O
even	O
in	O
activities	O
that	O
have	O
traditionally	O
been	O
considered	O
as	O
reserved	O
to	O
humans	O
,	O
like	O
painting	B-task
.	O

Modular	B-product
Audio	I-product
Recognition	I-product
Framework	I-product
(	O
MARF	B-product
)	O
is	O
an	O
open-source	O
research	O
platform	O
and	O
a	O
collection	O
of	O
voice	O
,	O
sound	O
,	O
speech	O
,	O
text	O
and	O
natural	B-field
language	I-field
processing	I-field
(	O
NLP	B-field
)	O
algorithm	O
s	O
written	O
in	O
Java	B-programlang
and	O
arranged	O
into	O
a	O
modular	O
and	O
extensible	O
framework	O
that	O
attempts	O
to	O
facilitate	O
addition	O
of	O
new	O
algorithm	O
s	O
.	O

In	O
2018	O
,	O
a	O
report	O
by	O
the	O
civil	O
liberties	O
and	O
rights	O
campaigning	O
organisation	O
Big	B-organisation
Brother	I-organisation
Watch	I-organisation
revealed	O
that	O
two	O
United	B-country
Kingdom	I-country
police	O
forces	O
,	O
South	B-organisation
Wales	I-organisation
Police	I-organisation
and	O
the	O
Metropolitan	B-organisation
Police	I-organisation
,	O
were	O
using	O
live	O
facial	B-task
recognition	I-task
at	O
public	O
events	O
and	O
in	O
public	O
spaces	O
,	O
in	O
September	O
2019	O
,	O
South	B-organisation
Wales	I-organisation
Police	I-organisation
use	O
of	O
facial	B-task
recognition	I-task
was	O
ruled	O
lawful	O
.	O

ANIMAL	B-product
has	O
been	O
ported	O
to	O
R	B-programlang
,	O
a	O
freely	O
available	O
language	O
and	O
environment	O
for	O
statistical	B-field
computing	I-field
and	O
graphics	B-field
.	O

Time-inhomogeneous	B-algorithm
hidden	I-algorithm
Bernoulli	I-algorithm
model	I-algorithm
(	O
TI-HBM	B-algorithm
)	O
is	O
an	O
alternative	O
to	O
hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
(	O
HMM	B-algorithm
)	O
for	O
automatic	B-task
speech	I-task
recognition	I-task
.	O

In	O
July	O
2016	O
,	O
Nvidia	B-organisation
demonstrated	O
during	O
SIGGRAPH	B-conference
a	O
new	O
method	O
of	O
foveated	O
rendering	O
claimed	O
to	O
be	O
invisible	O
to	O
users	O
.	O

Both	O
rely	O
on	O
speech	B-misc
act	I-misc
theory	I-misc
developed	O
by	O
John	B-researcher
Searle	I-researcher
in	O
the	O
1960s	O
and	O
enhanced	O
by	O
Terry	B-researcher
Winograd	I-researcher
and	O
Flores	B-researcher
in	O
the	O
1970s	O
.	O

Neural	B-algorithm
network	I-algorithm
models	I-algorithm
of	O
concept	O
formation	O
and	O
the	O
structure	O
of	O
knowledge	O
have	O
opened	O
powerful	O
hierarchical	O
models	O
of	O
knowledge	O
organization	O
such	O
as	O
George	B-researcher
Miller	I-researcher
'	O
s	O
Wordnet	B-product
.	O

Template	B-algorithm
matching	I-algorithm
has	O
various	O
applications	O
and	O
is	O
used	O
in	O
such	O
fields	O
as	O
face	B-task
recognition	I-task
(	O
see	O
facial	B-product
recognition	I-product
system	I-product
)	O
and	O
medical	B-task
image	I-task
processing	I-task
.	O

However	O
,	O
usage	O
only	O
became	O
widespread	O
in	O
2005	O
when	O
Navneet	B-researcher
Dalal	I-researcher
and	O
Bill	B-researcher
Triggs	I-researcher
,	O
researchers	O
for	O
the	O
French	B-organisation
National	I-organisation
Institute	I-organisation
for	I-organisation
Research	I-organisation
in	I-organisation
Computer	I-organisation
Science	I-organisation
and	I-organisation
Automation	I-organisation
(	O
INRIA	B-organisation
)	O
,	O
presented	O
their	O
supplementary	O
work	O
on	O
HOG	B-algorithm
descriptors	I-algorithm
at	O
the	O
Conference	B-conference
on	I-conference
Computer	I-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
(	O
CVPR	B-conference
)	O
.	O

Prior	O
to	O
joining	O
the	O
Penn	B-university
faculty	O
in	O
2002	O
,	O
he	O
spent	O
a	O
decade	O
(	O
1991-2001	O
)	O
in	O
AT	B-organisation
&	I-organisation
T	I-organisation
Labs	I-organisation
and	O
Bell	B-organisation
Labs	I-organisation
,	O
including	O
as	O
head	O
of	O
the	O
AI	B-field
department	O
with	O
colleagues	O
including	O
Michael	B-researcher
L.	I-researcher
Littman	I-researcher
,	O
David	B-researcher
A.	I-researcher
McAllester	I-researcher
,	O
and	O
Richard	B-researcher
S.	I-researcher
Sutton	I-researcher
;	O
Secure	B-organisation
Systems	I-organisation
Research	I-organisation
department	I-organisation
;	O
and	O
Machine	B-field
Learning	I-field
department	O
with	O
members	O
such	O
as	O
Michael	B-researcher
Collins	I-researcher
and	O
the	O
leader	O
)	O
.	O

When	O
data	O
are	O
unlabelled	O
,	O
supervised	B-field
learning	I-field
is	O
not	O
possible	O
,	O
and	O
an	O
unsupervised	B-field
learning	I-field
approach	O
is	O
required	O
which	O
attempts	O
to	O
find	O
natural	O
Cluster	B-task
analysis	I-task
to	O
groups	O
,	O
and	O
then	O
map	O
new	O
data	O
to	O
these	O
formed	O
groups	O
.	O

This	O
field	O
of	O
computer	B-field
science	I-field
developed	O
in	O
the	O
1950s	O
at	O
academic	O
institutions	O
such	O
as	O
the	O
MIT	B-organisation
A.I.	I-organisation
Lab	I-organisation
,	O
originally	O
as	O
a	O
branch	O
of	O
artificial	B-field
intelligence	I-field
and	O
robotics	B-field
.	O

It	O
could	O
also	O
be	O
replaced	O
by	O
the	O
Log	B-metrics
loss	I-metrics
equation	O
below	O
:	O

The	O
Shirley	B-organisation
Ryan	I-organisation
AbilityLab	I-organisation
(	O
formerly	O
the	O
Rehabilitation	B-organisation
Institute	I-organisation
of	I-organisation
Chicago	I-organisation
)	O
,	O
University	B-university
of	I-university
California	I-university
at	I-university
Berkeley	I-university
,	O
MIT	B-university
,	O
Stanford	B-university
University	I-university
,	O
and	O
University	B-university
of	I-university
Twente	I-university
in	O
the	O
Netherlands	B-country
are	O
the	O
researching	O
leaders	O
in	O
biomechatronics	B-field
.	O

Given	O
a	O
set	O
of	O
predicted	O
values	O
and	O
a	O
corresponding	O
set	O
of	O
actual	O
values	O
for	O
X	O
for	O
various	O
time	O
periods	O
,	O
a	O
common	O
evaluation	O
technique	O
is	O
to	O
use	O
the	O
mean	B-metrics
squared	I-metrics
prediction	I-metrics
error	I-metrics
;	O
other	O
measures	O
are	O
also	O
available	O
(	O
see	O
forecasting	O
#	O
forecasting	B-metrics
accuracy	I-metrics
)	O
.	O

Other	O
measures	O
,	O
such	O
as	O
the	O
proportion	O
of	O
correct	O
predictions	O
(	O
also	O
termed	O
accuracy	B-metrics
)	O
,	O
are	O
not	O
useful	O
when	O
the	O
two	O
classes	O
are	O
of	O
very	O
different	O
sizes	O
.	O

The	O
first	O
alpha	O
version	O
of	O
OpenCV	B-product
was	O
released	O
to	O
the	O
public	O
at	O
the	O
Conference	O
on	O
Computer	B-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
in	O
2000	O
,	O
and	O
five	O
betas	O
were	O
released	O
between	O
2001	O
and	O
2005	O
.	O

Results	O
have	O
been	O
presented	O
which	O
give	O
correlation	O
of	O
up	O
to	O
0.964	O
with	O
human	O
judgement	O
at	O
the	O
corpus	O
level	O
,	O
compared	O
to	O
BLEU	B-metrics
'	O
s	O
achievement	O
of	O
0.817	O
on	O
the	O
same	O
data	O
set	O
.	O

An	O
early	O
version	O
of	O
VMAF	B-metrics
has	O
been	O
shown	O
to	O
outperform	O
other	O
image	O
and	O
video	O
quality	O
metrics	O
such	O
as	O
SSIM	B-metrics
,	O
PSNR	B-metrics
-HVS	I-metrics
and	O
VQM-VFD	B-metrics
on	O
three	O
of	O
four	O
datasets	O
in	O
terms	O
of	O
prediction	O
accuracy	B-metrics
,	O
when	O
compared	O
to	O
subjective	O
ratings	O
.	O

For	O
example	O
,	O
the	O
ambiguity	O
of	O
'	O
mouse	O
'	O
(	O
animal	O
or	O
device	O
)	O
is	O
not	O
relevant	O
in	O
machine	B-task
translation	I-task
,	O
but	O
is	O
relevant	O
in	O
information	B-task
retrieval	I-task
.	O

Geometric	B-algorithm
hashing	I-algorithm
was	O
originally	O
suggested	O
in	O
computer	B-field
vision	I-field
for	O
object	B-task
recognition	I-task
in	O
2D	O
and	O
3D	O
,	O

It	O
forms	O
one	O
of	O
the	O
three	O
main	O
categories	O
of	O
machine	B-field
learning	I-field
,	O
along	O
with	O
supervised	B-field
learning	I-field
and	O
reinforcement	B-field
learning	I-field
.	O

Reinforcement	B-field
learning	I-field
,	O
due	O
to	O
its	O
generality	O
,	O
is	O
studied	O
in	O
many	O
other	O
disciplines	O
,	O
such	O
as	O
game	B-field
,	O
control	B-field
theory	I-field
,	O
operations	B-field
research	I-field
,	O
information	B-field
theory	I-field
,	O
simulation-based	B-field
optimization	I-field
,	O
multi-agent	B-field
systems	I-field
,	O
swarm	B-field
intelligence	I-field
,	O
statistics	B-field
and	O
genetic	B-algorithm
algorithm	I-algorithm
s	O
.	O

Pattern	B-field
recognition	I-field
is	O
closely	O
related	O
to	O
artificial	B-field
intelligence	I-field
and	O
machine	B-field
learning	I-field
,	O

The	O
software	O
is	O
used	O
to	O
design	O
,	O
train	O
and	O
deploy	O
neural	B-algorithm
network	I-algorithm
(	O
supervised	B-field
learning	I-field
and	O
unsupervised	B-field
learning	I-field
)	O
models	O
to	O
perform	O
a	O
wide	O
variety	O
of	O
tasks	O
such	O
as	O
data	B-field
mining	I-field
,	O
classification	B-task
,	O
function	B-task
approximation	I-task
,	O
multivariate	B-algorithm
regression	I-algorithm
and	O
time-series	B-task
prediction	I-task
.	O

In	O
2016	O
,	O
he	O
was	O
elected	O
Fellow	O
of	O
Association	B-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Artificial	I-conference
Intelligence	I-conference
.	O

She	O
serves	O
as	O
a	O
member	O
of	O
the	O
National	B-organisation
Academy	I-organisation
of	I-organisation
Sciences	I-organisation
(	O
since	O
2005	O
)	O
,	O
American	B-organisation
Academy	I-organisation
of	I-organisation
Arts	I-organisation
and	I-organisation
Sciences	I-organisation
(	O
since	O
2009	O
)	O
,	O

During	O
the	O
1973	O
Yom	B-misc
Kippur	I-misc
War	I-misc
,	O
Soviet-supplied	O
surface-to-air	B-product
missile	I-product
batteries	O
in	O
Egypt	B-country
and	O
Syria	B-country
caused	O
heavy	O
damage	O
Israeli	B-misc
fighter	O
jet	O
s	O
.	O

Another	O
resource	O
(	O
free	O
but	O
copyrighted	O
)	O
is	O
the	O
HTK	B-product
book	I-product
(	O
and	O
the	O
accompanying	O
HTK	B-product
toolkit	I-product
)	O
.	O

-	O
were	O
taken	O
in	O
the	O
2004	B-conference
AAAI	I-conference
Spring	O
Symposium	O
where	O
linguists	O
,	O
computer	O
scientists	O
,	O
and	O
other	O
interested	O
researchers	O
first	O
aligned	O
interests	O
and	O
proposed	O
shared	O
tasks	O
and	O
benchmark	O
data	O
sets	O
for	O
the	O
systematic	O
computational	O
research	O
on	O
affect	O
,	O
appeal	O
,	O
subjectivity	O
,	O
and	O
sentiment	O
in	O
text	O
.	O

A	O
single	O
grid	O
can	O
be	O
analysed	O
for	O
both	O
content	O
(	O
eyeball	B-task
inspection	I-task
)	O
and	O
structure	O
(	O
cluster	B-task
analysis	I-task
,	O
principal	B-task
component	I-task
analysis	I-task
,	O
and	O
a	O
variety	O
of	O
structural	O
indices	O
relating	O
to	O
the	O
complexity	O
and	O
range	O
of	O
the	O
ratings	O
being	O
the	O
chief	O
techniques	O
used	O
)	O
.	O

In	O
2018	O
Toyota	B-organisation
was	O
regarded	O
as	O
being	O
behind	O
in	O
Self-driving	B-product
car	I-product
and	O
in	O
need	O
of	O
innovation	O
.	O

Such	O
targets	O
include	O
natural	O
objects	O
such	O
as	O
ground	O
,	O
sea	O
,	O
precipitation	O
(	O
such	O
as	O
rain	O
,	O
snow	O
or	O
hail	O
)	O
,	O
sand	O
storm	O
s	O
,	O
animals	O
(	O
especially	O
birds	O
)	O
,	O
atmospheric	O
turbulence	O
,	O
and	O
other	O
atmospheric	O
effects	O
,	O
such	O
as	O
ionosphere	B-misc
reflections	I-misc
,	O
meteor	B-misc
trails	I-misc
,	O
and	O
three	B-misc
body	I-misc
scatter	I-misc
spike	I-misc
.	O

In	O
planning	O
and	O
control	O
,	O
the	O
essential	O
difference	O
between	O
humanoids	O
and	O
other	O
kinds	O
of	O
robots	O
(	O
like	O
industrial	B-product
ones	O
)	O
is	O
that	O
the	O
movement	O
of	O
the	O
robot	O
must	O
be	O
human-like	O
,	O
using	O
legged	O
locomotion	O
,	O
especially	O
biped	B-misc
gait	I-misc
.	O

The	O
gradient	B-algorithm
descent	I-algorithm
can	O
take	O
many	O
iterations	O
to	O
compute	O
a	O
local	B-misc
minimum	I-misc
with	O
a	O
required	O
accuracy	B-metrics
,	O
if	O
the	O
curvature	B-misc
in	O
different	O
directions	O
is	O
very	O
different	O
for	O
the	O
given	O
function	O
.	O

The	O
1997	B-misc
RoboCup	I-misc
2D	I-misc
Soccer	I-misc
Simulation	I-misc
League	I-misc
was	O
the	O
first	O
RoboCup	B-misc
competition	O
promoted	O
in	O
conjunction	O
with	O
International	B-conference
Joint	I-conference
Conference	I-conference
on	I-conference
Artificial	I-conference
Intelligence	I-conference
held	O
in	O
Nagoya	B-location
,	O
Japan	B-country
,	O
from	O
23	O
to	O
29	O
August	O
1997	O
.	O

Other	O
programming	O
options	O
include	O
an	O
embedded	O
Python	B-programlang
environment	O
,	O
and	O
an	O
R	B-programlang
Console	O
plus	O
support	O
for	O
Rserve	B-product
.	O

From	O
Bonn	B-researcher
he	O
has	O
contributed	O
fundamentally	O
to	O
artificial	B-field
intelligence	I-field
and	O
robotics	B-field
(	O
with	O
Wolfram	B-researcher
Burgard	I-researcher
,	O
Dieter	B-researcher
Fox	I-researcher
,	O
Sebastian	B-researcher
Thrun	I-researcher
among	O
his	O
students	O
)	O
,	O
and	O
to	O
the	O
development	O
of	O
software	B-field
engineering	I-field
,	O
particularly	O
in	O
civil	B-field
engineering	I-field
,	O
and	O
information	B-field
systems	I-field
,	O
particularly	O
in	O
the	O
geosciences.	B-field
won	O
the	O
AAAI	B-misc
Classic	I-misc
Paper	I-misc
award	I-misc
of	O
2016.2014	O
.	O

The	O
first	O
USA	B-conference
edition	I-conference
of	I-conference
Campus	I-conference
Party	I-conference
will	O
take	O
place	O
from	O
20	O
to	O
22	O
of	O
August	O
at	O
TCF	B-location
Center	I-location
in	O
Detroit	B-location
,	O
Michigan	B-location
.	O

Together	O
with	O
Yann	B-researcher
LeCun	I-researcher
,	O
and	O
Yoshua	B-researcher
Bengio	I-researcher
,	O
Hinton	B-researcher
won	O
the	O
2018	O
Turing	B-misc
Award	I-misc
for	O
conceptual	O
and	O
engineering	O
breakthroughs	O
that	O
have	O
made	O
deep	B-algorithm
neural	I-algorithm
networks	I-algorithm
a	O
critical	O
component	O
of	O
computing	O
.	O

Euler	B-product
Math	I-product
Toolbox	I-product
uses	O
a	O
matrix	O
language	O
similar	O
to	O
MATLAB	B-product
,	O
a	O
system	O
that	O
had	O
been	O
under	O
development	O
since	O
the	O
1970s	O
.	O

Some	O
languages	O
make	O
it	O
possible	O
portably	O
(	O
e.g.	O
Scheme	B-programlang
,	O
Common	B-programlang
Lisp	I-programlang
,	O
Perl	B-programlang
or	O
D	B-programlang
)	O
.	O

In	O
1969	O
a	O
famous	O
book	O
entitled	O
Perceptrons	B-misc
by	O
Marvin	B-researcher
Minsky	I-researcher
and	O
Seymour	B-researcher
Papert	I-researcher
showed	O
that	O
it	O
was	O
impossible	O
for	O
these	O
classes	O
of	O
network	O
to	O
learn	O
an	O
XOR	B-misc
function	I-misc
.	O

Large	O
numbers	O
of	O
Russian	B-misc
scientific	O
and	O
technical	O
documents	O
were	O
translated	O
using	O
SYSTRAN	B-product
under	O
the	O
auspices	O
of	O
the	O
USAF	B-organisation
Foreign	I-organisation
Technology	I-organisation
Division	I-organisation
(	O
later	O
the	O
National	B-organisation
Air	I-organisation
and	I-organisation
Space	I-organisation
Intelligence	I-organisation
Center	I-organisation
)	O
at	O
Wright-Patterson	B-location
Air	I-location
Force	I-location
Base	I-location
,	O
Ohio	B-location
.	O

Semi-supervised	B-field
learning	I-field
falls	O
between	O
unsupervised	B-field
learning	I-field
(	O
without	O
any	O
labeled	O
training	O
data	O
)	O
and	O
supervised	B-field
learning	I-field
(	O
with	O
completely	O
labeled	O
training	O
data	O
)	O
.	O

An	O
n	B-algorithm
-gram	I-algorithm
model	I-algorithm
is	O
a	O
type	O
of	O
probabilistic	B-algorithm
language	I-algorithm
model	I-algorithm
for	O
predicting	O
the	O
next	O
item	O
in	O
such	O
a	O
sequence	O
in	O
the	O
form	O
of	O
a	O
(	O
n	O
−	O
1	O
)	O
-order	O
Markov	B-algorithm
model	I-algorithm
.efficiently	O
.	O

The	O
Cleveland	B-location
Clinic	I-location
has	O
used	O
Cyc	B-product
to	O
develop	O
a	O
natural	B-product
language	I-product
query	I-product
interface	I-product
of	I-product
biomedical	I-product
information	I-product
,	O
spanning	O
decades	O
of	O
information	O
on	O
cardiothoracic	O
surgeries	O
.	O

The	O
incident	O
strained	O
relations	O
between	O
the	O
United	B-country
States	I-country
and	O
Japan	B-country
,	O
and	O
resulted	O
in	O
the	O
arrest	O
and	O
prosecution	O
two	O
senior	O
executives	O
,	O
as	O
well	O
as	O
the	O
imposition	O
of	O
sanctions	O
on	O
the	O
company	O
by	O
both	O
countries	O
.	O

If	O
the	O
modeling	O
is	O
done	O
by	O
an	O
artificial	B-algorithm
neural	I-algorithm
network	I-algorithm
or	O
other	O
machine	B-field
learning	I-field
,	O
the	O
optimization	O
of	O
parameters	O
is	O
called	O
training	B-misc
,	O
while	O
the	O
optimization	O
of	O
model	O
hyperparameters	O
is	O
called	O
tuning	B-misc
and	O
often	O
uses	O
cross-validation	B-algorithm
..	O

Localized	O
versions	O
of	O
the	O
site	O
available	O
in	O
the	O
United	B-country
Kingdom	I-country
,	O
India	B-country
,	O
and	O
Australia	B-country
were	O
discontinued	O
following	O
the	O
acquisition	O
of	O
Rotten	B-organisation
Tomatoes	I-organisation
by	O
Fandango	B-organisation
.	O

The	O
NER	B-task
model	O
is	O
one	O
of	O
a	O
number	O
of	O
methods	O
for	O
determining	O
the	O
accuracy	B-metrics
of	O
live	O
subtitles	O
in	O
television	O
broadcasts	O
and	O
events	O
that	O
are	O
produced	O
using	O
speech	B-task
recognition	I-task
.	O

Atran	B-researcher
has	O
taught	O
at	O
Cambridge	B-university
University	I-university
,	O
Hebrew	B-university
University	I-university
in	O
Jerusalem	B-location
,	O
the	O
École	B-university
pratique	I-university
des	I-university
hautes	I-university
études	I-university
and	O
École	B-university
Polytechnique	I-university
in	O
Paris	B-location
,	O
and	O
John	B-university
Jay	I-university
College	I-university
of	I-university
Criminal	I-university
Justice	I-university
in	O
New	B-location
York	I-location
City	I-location
.	O

SHRDLU	B-product
was	O
an	O
early	O
natural	B-task
language	I-task
understanding	I-task
computer	O
program	O
,	O
developed	O
by	O
Terry	B-researcher
Winograd	I-researcher
at	O
MIT	B-university
in	O
1968-1970	O

He	O
received	O
a	O
B.E.	B-misc
in	O
electronics	B-field
engineering	I-field
from	O
B.M.S.	B-university
College	I-university
of	I-university
Engineering	I-university
in	O
Bangalore	B-location
,	O
India	B-country
in	O
1982	O
,	O
when	O
it	O
was	O
affiliated	O
with	O
Bangalore	B-university
University	I-university
,	O
an	O
M.S.	B-misc
in	O
electrical	B-field
and	I-field
computer	I-field
engineering	I-field
in	O
1984	O
from	O
Drexel	B-university
University	I-university
,	O
and	O
an	O
M.S.	B-misc
in	O
computer	B-field
science	I-field
in	O
1989	O
,	O
and	O
a	O
Ph.D.	B-misc
in	O
1990	O
,	O
respectively	O
,	O
from	O
the	O
University	B-university
of	I-university
Wisconsin-Madison	I-university
,	O
where	O
he	O
studied	O
Artificial	B-field
Intelligence	I-field
and	O
worked	O
with	O
Leonard	B-researcher
Uhr	I-researcher
.	O

Accuracy	O
is	O
usually	O
rated	O
with	O
word	B-metrics
error	I-metrics
rate	I-metrics
(	O
WER	B-metrics
)	O
,	O
whereas	O
speed	O
is	O
measured	O
with	O
the	O
real	B-metrics
time	I-metrics
factor	I-metrics
.	O

In	O
1971	O
Terry	B-researcher
Winograd	I-researcher
developed	O
an	O
early	O
natural	B-field
language	I-field
processing	I-field
engine	O
capable	O
of	O
interpreting	O
naturally	O
written	O
commands	O
within	O
a	O
simple	O
rule-governed	O
environment	O
.	O

In	O
artificial	B-field
intelligence	I-field
,	O
Marvin	B-researcher
Minsky	I-researcher
,	O
Herbert	B-researcher
A.	I-researcher
Simon	I-researcher
,	O
and	O
Allen	B-researcher
Newell	I-researcher
are	O
prominent	O
.	O

In	O
the	O
latter	O
half	O
of	O
the	O
20th	O
century	O
,	O
electrical	B-field
engineering	I-field
itself	O
separated	O
into	O
several	O
disciplines	O
,	O
specialising	O
in	O
the	O
design	O
and	O
analysis	O
of	O
systems	O
that	O
manipulate	O
physical	O
signals	O
;	O
electronic	B-field
engineering	I-field
and	O
computer	B-field
engineering	I-field
as	O
examples	O
;	O
while	O
design	B-field
engineering	I-field
developed	O
to	O
deal	O
with	O
functional	O
design	O
of	O
user-machine	B-misc
interfaces	I-misc
.	O

Perhaps	O
the	O
simplest	O
statistic	O
is	O
accuracy	B-metrics
or	O
Fraction	B-metrics
Correct	I-metrics
(	O
FC	B-metrics
)	O
,	O
which	O
measures	O
the	O
fraction	O
of	O
all	O
instances	O
that	O
are	O
correctly	O
categorized	O
;	O
it	O
is	O
the	O
ratio	O
of	O
the	O
number	O
of	O
correct	O
classifications	O
to	O
the	O
total	O
number	O
of	O
correct	O
or	O
incorrect	O
classifications	O
:	O
(	O
TP	B-metrics
+	I-metrics
TN	I-metrics
)	O
/	O
Total	O
Population	O
=	O
(	O
TP	B-metrics
+	I-metrics
TN	I-metrics
)	O
/	O
(	O
TP	B-metrics
+	I-metrics
TN	I-metrics
+	I-metrics
FP	I-metrics
+	I-metrics
FN	I-metrics
)	O
.	O

In	O
the	O
academic	O
community	O
,	O
the	O
major	O
forums	O
for	O
research	O
started	O
in	O
1995	O
when	O
the	O
First	B-conference
International	I-conference
Conference	I-conference
Data	I-conference
Mining	I-conference
and	I-conference
Knowledge	I-conference
Discovery	I-conference
(	O
KDD-95	B-conference
)	O
was	O
started	O
in	O
Montreal	B-location
under	O
AAAI	B-conference
sponsorship	O
.	O

In	O
this	O
approach	O
,	O
models	O
are	O
developed	O
using	O
different	O
data	B-field
mining	I-field
,	O
machine	B-field
learning	I-field
algorithms	O
to	O
predict	O
users	O
'	O
rating	O
of	O
unrated	O
items	O
.	O

In	O
light	O
of	O
the	O
above	O
discussion	O
,	O
we	O
see	O
that	O
the	O
SVM	B-algorithm
technique	O
is	O
equivalent	O
to	O
empirical	B-algorithm
risk	I-algorithm
with	O
Tikhonov	B-algorithm
regularization	I-algorithm
,	O
where	O
in	O
this	O
case	O
the	O
loss	B-misc
function	I-misc
is	O
the	O
hinge	B-metrics
loss	I-metrics

The	O
2015	O
edition	O
was	O
hosted	O
by	O
Molly	B-person
McGrath	I-person
,	O
with	O
Chris	B-person
Rose	I-person
and	O
former	O
UFC	B-organisation
fighter	O
Kenny	B-person
Florian	I-person
as	O
commentators	O
.	O

A	O
subset	O
called	O
Micro-Planner	B-product
was	O
implemented	O
by	O
Gerald	B-researcher
Jay	I-researcher
Sussman	I-researcher
,	O
Eugene	B-researcher
Charniak	I-researcher
and	O
Terry	B-researcher
Winograd	I-researcher
Sussman	B-researcher
,	O
,	O
and	O
Winograd	B-researcher
1971	O
and	O
was	O
used	O
in	O
Winograd	B-researcher
's	O
natural-language	B-task
understanding	I-task
program	O
SHRDLU	B-product
,	O
Eugene	B-researcher
Charniak	I-researcher
's	O
story	B-task
understanding	I-task
work	O
,	O
Thorne	B-researcher
McCarty	I-researcher
's	O
work	O
on	O
legal	B-task
reasoning	I-task
,	O
and	O
some	O
other	O
projects	O
.	O

WordNet	B-product
has	O
been	O
used	O
for	O
a	O
number	O
of	O
purposes	O
in	O
information	B-product
systems	I-product
,	O
including	O
word-sense	B-task
disambiguation	I-task
,	O
information	B-task
retrieval	I-task
,	O
automatic	B-task
text	I-task
classification	I-task
,	O
Automatic	B-task
summarization	I-task
,	O
machine	B-task
translation	I-task
and	O
even	O
automatic	B-task
crossword	I-task
puzzle	I-task
generation	I-task
.	O

Keutzer	B-researcher
was	O
named	O
a	O
Fellow	O
of	O
the	O
IEEE	B-organisation
in	O
1996	O
.	O

A	O
widely	O
used	O
type	O
of	O
composition	O
is	O
the	O
nonlinear	B-algorithm
weighted	I-algorithm
sum	I-algorithm
,	O
where	O
math	O
\	O
textstyle	O
f	O
(	O
x	O
)	O
=	O
K	O
\	O
left	O
(	O
\	O
sum	O
_	O
i	O
w	O
_	O
i	O
g	O
_	O
i	O
(	O
x	O
)	O
\	O
right	O
)	O
/	O
math	O
,	O
where	O
math	O
\	O
textstyle	O
K	O
/	O
math	O
(	O
commonly	O
referred	O
to	O
as	O
the	O
activation	B-misc
function	I-misc
)	O
is	O
some	O
predefined	O
function	O
,	O
such	O
as	O
the	O
hyperbolic	B-algorithm
tangent	I-algorithm
,	O
sigmoid	B-algorithm
function	I-algorithm
,	O
softmax	B-algorithm
function	I-algorithm
,	O
or	O
rectifier	B-algorithm
function	I-algorithm
.	O

In	O
the	O
film	O
Westworld	B-misc
,	O
female	O
robots	O
actually	O
engaged	O
in	O
intercourse	O
with	O
human	O
men	O
as	O
part	O
of	O
the	O
make-believe	O
vacation	O
world	O
human	O
customers	O
paid	O
to	O
attend	O
.	O

Typically	O
,	O
the	O
process	O
starts	O
by	O
terminology	B-task
extraction	I-task
and	O
concepts	O
or	O
noun	O
phrase	O
s	O
from	O
plain	O
text	O
using	O
linguistic	O
processors	O
such	O
as	O
part-of-speech	B-task
tagging	I-task
and	O
phrase	B-task
chunking	I-task
.	O

They	O
demonstrated	O
its	O
performance	O
on	O
a	O
number	O
of	O
problems	O
of	O
interest	O
to	O
the	O
machine	B-field
learning	I-field
community	O
,	O
including	O
handwriting	B-task
recognition	I-task
.	O

While	O
studying	O
at	O
Stanford	B-university
,	O
Scheinman	B-researcher
was	O
awarded	O
a	O
fellowship	O
sponsored	O
by	O
George	B-researcher
Devol	I-researcher
,	O
the	O
inventor	O
of	O
the	O
Unimate	B-product
,	O
the	O
first	O
industrial	B-product
robot	I-product
.	O

While	O
originally	O
used	O
to	O
evaluate	O
machine	B-task
translations	I-task
,	O
bilingual	B-metrics
evaluation	I-metrics
understudy	I-metrics
(	O
BLEU	B-metrics
)	O
has	O
been	O
used	O
successfully	O
to	O
evaluate	O
paraphrase	B-product
generation	I-product
models	I-product
as	O
well	O
.	O

Unimation	B-organisation
later	O
licensed	O
their	O
technology	O
to	O
Kawasaki	B-organisation
Heavy	I-organisation
Industries	I-organisation
and	O
GKN	B-organisation
,	O
manufacturing	O
Unimate	B-product
s	O
in	O
Japan	B-country
and	O
England	B-country
respectively	O
.	O

Much	O
of	O
the	O
confusion	O
between	O
these	O
two	O
research	O
communities	O
(	O
which	O
do	O
often	O
have	O
separate	O
conferences	O
and	O
separate	O
journals	O
,	O
ECML	B-conference
PKDD	I-conference
being	O
a	O
major	O
exception	O
)	O
comes	O
from	O
the	O
basic	O
assumptions	O
they	O
work	O
with	O
:	O
in	O
machine	B-field
learning	I-field
,	O
performance	O
is	O
usually	O
evaluated	O
with	O
respect	O
to	O
the	O
ability	O
to	O
reproduce	O
known	O
knowledge	O
,	O
while	O
in	O
knowledge	B-conference
discovery	I-conference
and	I-conference
data	I-conference
mining	I-conference
(	O
KDD	B-conference
)	O
the	O
key	O
task	O
is	O
the	O
discovery	O
of	O
previously	O
unknown	O
knowledge	O
.	O

Hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
s	O
are	O
the	O
basis	O
for	O
most	O
modern	O
automatic	B-product
speech	I-product
recognition	I-product
systems	I-product
.	O

,	O
a	O
company	O
in	O
Bangalore	B-location
,	O
India	B-country
specializing	O
in	O
online	O
handwriting	B-task
recognition	I-task
software	O
.	O

Do	O
repeated	O
translations	O
converge	O
on	O
a	O
single	O
expression	O
in	O
both	O
languages	O
?	O
I.e.	O
does	O
the	O
translation	O
method	O
show	O
stationarity	O
or	O
produce	O
a	O
canonical	B-misc
form	I-misc
?	O
Does	O
the	O
translation	O
become	O
stationary	O
without	O
losing	O
the	O
original	O
meaning	O
?	O
This	O
metric	O
has	O
been	O
criticized	O
as	O
not	O
being	O
well	O
correlated	O
with	O
BLEU	B-metrics
(	O
BiLingual	B-metrics
Evaluation	I-metrics
Understudy	I-metrics
)	O
scores	O
.	O

He	O
holds	O
fellowships	O
in	O
the	O
American	B-conference
Association	I-conference
for	I-conference
Artificial	I-conference
Intelligence	I-conference
,	O
the	O
Center	B-organisation
for	I-organisation
Advanced	I-organisation
Study	I-organisation
in	I-organisation
the	I-organisation
Behavioral	I-organisation
Sciences	I-organisation
at	O
Stanford	B-university
University	I-university
,	O
the	O
MIT	B-university
Center	O
for	O
Cognitive	B-field
Science	I-field
,	O
the	O
Canadian	B-organisation
Institute	I-organisation
for	I-organisation
Advanced	I-organisation
Research	I-organisation
,	O
the	O
Canadian	B-organisation
Psychological	I-organisation
Association	I-organisation
,	O
and	O
was	O
elected	O
Fellow	O
of	O
the	O
Royal	B-organisation
Society	I-organisation
of	I-organisation
Canada	I-organisation
in	O
1998	O
.	O

Hinton	B-researcher
-	O
together	O
with	O
Yoshua	B-researcher
Bengio	I-researcher
and	O
Yann	B-researcher
LeCun	I-researcher
-	O
are	O
referred	O
to	O
by	O
some	O
as	O
the	O
Godfathers	B-misc
of	I-misc
AI	I-misc
and	O
Godfathers	B-misc
of	I-misc
Deep	I-misc
Learning	I-misc
.	O

The	O
lightweight	O
open-source	O
speech	O
project	O
eSpeak	B-product
,	O
which	O
has	O
its	O
own	O
approach	O
to	O
synthesis	O
,	O
has	O
experimented	O
with	O
Mandarin	B-misc
and	O
Cantonese.	B-misc
eSpeak	B-product
was	O
used	O
by	O
Google	B-product
Translate	I-product
from	O
May	O
20102010	O
.	O

Also	O
released	O
in	O
1982	O
,	O
Software	B-product
Automatic	I-product
Mouth	I-product
was	O
the	O
first	O
commercial	O
all-software	O
voice	O
synthesis	B-task
program	I-task
.	O

The	O
column	O
ratios	O
are	O
TRUE	B-metrics
Positive	I-metrics
Rate	I-metrics
(	O
TPR	B-metrics
,	O
aka	O
Sensitivity	B-metrics
or	O
recall	B-metrics
)	O
(	O
TP	B-metrics
/	I-metrics
(	I-metrics
TP	I-metrics
+	I-metrics
FN	I-metrics
)	I-metrics
)	O
,	O
with	O
complement	O
the	O
FALSE	B-metrics
Negative	I-metrics
Rate	I-metrics
(	O
FNR	B-metrics
)	O
(	O
FN	B-metrics
/	I-metrics
(	I-metrics
TP	I-metrics
+	I-metrics
FN	I-metrics
)	I-metrics
)	O
;	O
and	O
TRUE	B-metrics
Negative	I-metrics
Rate	I-metrics
(	O
TNR	B-metrics
,	O
aka	O
Specificity	B-metrics
,	O
SPC	B-metrics
)	O
(	O
TN	B-metrics
/	I-metrics
(	I-metrics
TN	I-metrics
+	I-metrics
FP	I-metrics
)	I-metrics
)	O
,	O
with	O
complement	O
FALSE	B-metrics
Positive	I-metrics
Rate	I-metrics
(	O
FPR	B-metrics
)	O
(	O
FP	B-metrics
/	I-metrics
(	I-metrics
TN	I-metrics
+	I-metrics
FP	I-metrics
)	I-metrics
)	O
.	O

Edsinger	B-person
and	O
Weber	B-organisation
collaborated	O
on	O
many	O
other	O
robots	O
as	O
well	O
,	O
and	O
their	O
experience	O
working	O
with	O
the	O
Kismet	B-product

R	B-programlang
functionality	O
is	O
accessible	O
from	O
several	O
scripting	O
languages	O
such	O
as	O
Python	B-programlang
,	O
are	O
available	O
as	O
well	O
.	O

VAL	B-programlang
was	O
one	O
of	O
the	O
first	O
robot	O
languages	O
and	O
was	O
used	O
in	O
Unimate	B-product
robots	I-product
.	O

They	O
presented	O
their	O
database	O
for	O
the	O
first	O
time	O
as	O
a	O
poster	O
at	O
the	O
2009	B-conference
Conference	I-conference
on	I-conference
Computer	I-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
(	O
CVPR	B-conference
)	O
in	O
Florida	B-location
.	O

Categorization	B-misc
tasks	I-misc
in	O
which	O
no	O
labels	O
are	O
supplied	O
are	O
referred	O
to	O
as	O
unsupervised	B-task
classification	I-task
,	O
unsupervised	B-field
learning	I-field
,	O
Cluster	B-task
analysis	I-task
.	O

It	O
needs	O
to	O
Object	B-task
recognition	I-task
,	O
recognize	O
and	O
locate	O
humans	O
and	O
further	O
emotion	B-task
recognition	I-task
.	O

The	O
process	O
is	O
complex	O
and	O
contains	O
encoding	B-misc
and	O
recall	B-misc
or	O
retrieval	B-misc
.	O

Also	O
known	O
as	O
parallel	O
robots	O
,	O
or	O
generalized	O
Stewart	B-product
platforms	I-product
(	O
in	O
the	O
Stewart	B-product
platform	I-product
,	O
the	O
actuators	O
are	O
paired	O
together	O
on	O
both	O
the	O
basis	O
and	O
the	O
platform	O
)	O
,	O
these	O
systems	O
are	O
articulated	B-product
robot	I-product
s	O
that	O
use	O
similar	O
mechanisms	O
for	O
the	O
movement	O
of	O
either	O
the	O
robot	O
on	O
its	O
base	O
,	O
or	O
one	O
or	O
more	O
manipulator	O
arms	O
.	O

Machine	B-field
vision	I-field
as	O
a	O
systems	B-field
engineering	I-field
discipline	O
can	O
be	O
considered	O
distinct	O
from	O
computer	B-field
vision	I-field
,	O
a	O
form	O
of	O
computer	B-field
science	I-field
.	O

The	O
activation	O
function	O
of	O
the	O
LSTM	B-algorithm
gates	I-algorithm
is	O
often	O
the	O
logistic	B-algorithm
sigmoid	I-algorithm
function	I-algorithm
.	O

In	O
other	O
words	O
,	O
the	O
sample	B-metrics
mean	I-metrics
is	O
the	O
(	O
necessarily	O
unique	O
)	O
efficient	O
estimator	O
,	O
and	O
thus	O
also	O
the	O
minimum	B-metrics
variance	I-metrics
unbiased	I-metrics
estimator	I-metrics
(	O
MVUE	B-metrics
)	O
,	O
in	O
addition	O
to	O
being	O
the	O
maximum	B-metrics
likelihood	I-metrics
estimator	I-metrics
.	O

The	O
2001	O
Scientific	B-misc
American	I-misc
article	O
by	O
Berners-Lee	B-researcher
,	O
James	B-researcher
Hendler	I-researcher
,	O
and	O
Ora	B-researcher
Lassila	I-researcher
described	O
an	O
expected	O
evolution	O
of	O
the	O
existing	O
Web	B-product
to	O
a	O
Semantic	B-product
Web	I-product
.	O

Blade	B-misc
Runner	I-misc
used	O
a	O
number	O
of	O
then-lesser-known	O
actors	O
:	O
Sean	B-person
Young	I-person
portrays	O
Rachael	B-person
,	O
an	O
experimental	O
replicant	O
implanted	O
with	O
the	O
memories	O
of	O
Tyrell	B-person
's	O
niece	O
,	O
causing	O
her	O
to	O
believe	O
she	O
is	O
human	O
;	O
Sammon	B-person
,	O
pp.	O
92-93	O
Nina	B-person
Axelrod	I-person
auditioned	O
for	O
the	O
role	O
.	O

Gerry	B-researcher
Sussman	I-researcher
,	O
Eugene	B-researcher
Charniak	I-researcher
,	O
Seymour	B-researcher
Papert	I-researcher
and	O
Terry	B-researcher
Winograd	I-researcher
visited	O
the	O
University	B-university
of	I-university
Edinburgh	I-university
in	O
1971	O
spreading	O
the	O
news	O
about	O
Micro-Planner	B-product
and	O
SHRDLU	B-product
and	O
casting	O
doubt	O
on	O
the	O
resolution	O
uniform	O
proof	O
procedure	O
approach	O
that	O
had	O
been	O
the	O
mainstay	O
of	O
the	O
Edinburgh	B-location
Logicists	O
.	O

Walter	B-researcher
's	O
work	O
inspired	O
subsequent	O
generations	O
of	O
robotics	B-field
researchers	O
such	O
as	O
Rodney	B-researcher
Brooks	I-researcher
,	O
Hans	B-researcher
Moravec	I-researcher
and	O
Mark	B-researcher
Tilden	I-researcher
.	O

Subsequently	O
,	O
a	O
similar	O
GPU-based	O
CNN	B-algorithm
by	O
Alex	B-researcher
Krizhevsky	I-researcher
et	O
al.	O
won	O
the	O
ImageNet	B-conference
Large	I-conference
Scale	I-conference
Visual	I-conference
Recognition	I-conference
Challenge	I-conference
2012	I-conference
.	O

Commonly	O
used	O
loss	B-misc
functions	I-misc
for	O
probabilistic	O
classification	O
include	O
log	B-metrics
loss	I-metrics
and	O
the	O
Brier	B-metrics
score	I-metrics
between	O
the	O
predicted	O
and	O
the	O
TRUE	B-misc
probability	I-misc
distributions	O
.	O

In	O
May	O
2016	O
,	O
NtechLab	B-organisation
was	O
admitted	O
to	O
the	O
official	O
testing	O
of	O
biometrics	B-field
technology	O
by	O
NIST	B-organisation
among	O
the	O
three	O
Russian	B-misc
companies	O
.	O

However	O
,	O
floating-point	O
numbers	O
have	O
only	O
a	O
certain	O
amount	O
of	O
mathematical	O
precision	O
.	O

During	O
2015	O
,	O
many	O
of	O
SenseTime	B-organisation
's	O
papers	O
were	O
accepted	O
into	O
the	O
Conference	B-conference
on	I-conference
Computer	I-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
(	O
CVPR	B-conference
)	O
.	O

He	O
co-developed	O
optimal	O
algorithms	O
for	O
Structure	B-task
From	I-task
Motion	I-task
(	O
SFM	B-task
,	O
or	O
Visual	B-task
SLAM	I-task
,	O
simultaneous	B-task
localization	I-task
and	I-task
mapping	I-task
,	O
in	O
Robotics	B-field
;	O
Best	B-misc
Paper	I-misc
Award	I-misc
at	O
Conference	B-conference
on	I-conference
Computer	I-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
1998	O
)	O
,	O
characterized	O
its	O
ambiguities	O
(	O
David	B-misc
Marr	I-misc
Prize	I-misc
at	O
ICCV	B-conference
1999	I-conference
)	O
,	O
also	O
characterized	O
the	O
identifiability	O
and	O
observability	O
of	O
visual-inertial	O
sensor	O
fusion	O
(	O
Best	B-misc
Paper	I-misc
Award	I-misc
at	O
Robotics	B-field
2015	O
)	O
.	O

Stephen	B-researcher
H.	I-researcher
Muggleton	I-researcher
FBCS	B-organisation
,	O
FIET	B-organisation
,	O
Association	B-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Artificial	I-conference
Intelligence	I-conference
,	O

Edge	B-task
detection	I-task
is	O
a	O
fundamental	O
tool	O
in	O
image	B-field
processing	I-field
,	O
machine	B-field
vision	I-field
and	O
computer	B-field
vision	I-field
,	O
particularly	O
in	O
the	O
areas	O
of	O
feature	B-task
detection	I-task
and	O
feature	B-task
extraction	I-task
.	O

An	O
example	O
of	O
this	O
would	O
be	O
a	O
variable	O
such	O
as	O
outside	B-misc
temperature	I-misc
(	O
mathtemp	O
/	O
math	O
)	O
,	O
which	O
in	O
a	O
given	O
application	O
might	O
be	O
recorded	O
to	O
several	O
decimal	B-misc
places	I-misc
of	I-misc
precision	I-misc
(	O
depending	O
on	O
the	O
sensing	O
apparatus	O
)	O
.	O

The	O
returning	O
judges	O
are	O
Fon	B-person
Davis	I-person
,	O
Jessica	B-person
Chobot	I-person
,	O
and	O
Leland	B-person
Melvin	I-person
,	O
as	O
well	O
as	O
celebrity	O
guest	O
judges	O
actor	O
Clark	B-person
Gregg	I-person
,	O
MythBusters	B-misc
host	O
and	O
former	O
Battlebots	B-misc
builder	O
Adam	B-person
Savage	I-person
,	O
NFL	B-organisation
tightend	O
Vernon	B-person
Davis	I-person
,	O
and	O
YouTube	B-organisation
star	O
Michael	B-person
Stevens	I-person
a.k.a.	O
Vsauce	B-person
.	O

But	O
these	O
methods	O
never	O
won	O
over	O
the	O
non-uniform	O
internal-handcrafting	O
Gaussian	B-algorithm
mixture	I-algorithm
model	I-algorithm
/	O
Hidden	B-algorithm
Markov	I-algorithm
model	I-algorithm
(	O
GMM-HMM	B-algorithm
)	O
technology	O
based	O
on	O
generative	O
models	O
of	O
speech	O
trained	O
discriminatively	O
.	O

Software	O
packages	O
like	O
MATLAB	B-product
,	O
GNU	B-programlang
Octave	I-programlang
,	O
Scilab	B-programlang
,	O
and	O
SciPy	B-product
provide	O
convenient	O
ways	O
to	O
apply	O
these	O
different	O
methods	O
.	O

Linear	B-algorithm
predictive	I-algorithm
coding	I-algorithm
(	O
LPC	B-algorithm
)	O
,	O
a	O
speech	B-task
processing	I-task
algorithm	O
,	O
was	O
first	O
proposed	O
by	O
Fumitada	B-researcher
Itakura	I-researcher
of	O
Nagoya	B-university
University	I-university
and	O
Shuzo	B-researcher
Saito	I-researcher
of	O
Nippon	B-organisation
Telegraph	I-organisation
and	I-organisation
Telephone	I-organisation
(	O
NTT	B-organisation
)	O
in	O
1966	O
.	O

In	O
2006	O
,	O
for	O
the	O
25th	O
anniversary	O
of	O
the	O
algorithm	O
,	O
a	O
workshop	O
was	O
organized	O
at	O
the	O
International	B-conference
Conference	I-conference
on	I-conference
Computer	I-conference
Vision	I-conference
and	I-conference
Pattern	I-conference
Recognition	I-conference
(	O
CVPR	B-conference
)	O
to	O
summarize	O
the	O
most	O
recent	O
contributions	O
and	O
variations	O
to	O
the	O
original	O
algorithm	O
,	O
mostly	O
meant	O
to	O
improve	O
the	O
speed	O
of	O
the	O
algorithm	O
,	O
the	O
robustness	O
and	O
accuracy	O
of	O
the	O
estimated	O
solution	O
and	O
to	O
decrease	O
the	O
dependency	O
from	O
user	O
defined	O
constants	O
.	O

The	O
members	O
went	O
to	O
the	O
University	B-university
of	I-university
Debrecen	I-university
,	O
the	O
Hungarian	B-organisation
Academy	I-organisation
of	I-organisation
Sciences	I-organisation
,	O
Eötvös	B-university
Loránd	I-university
University	I-university
,	O
etc	O
.	O

To	O
extend	O
SVM	B-algorithm
to	O
cases	O
in	O
which	O
the	O
data	O
are	O
not	O
linearly	O
separable	O
,	O
we	O
introduce	O
the	O
loss	B-misc
function	I-misc
,	O

Logo	B-programlang
is	O
an	O
educational	O
programming	O
language	O
,	O
designed	O
in	O
1967	O
by	O
Wally	B-researcher
Feurzeig	I-researcher
,	O
Seymour	B-researcher
Papert	I-researcher
,	O
and	O
Cynthia	B-researcher
Solomon	I-researcher
.	O

Eyring	B-organisation
Research	I-organisation
Institute	I-organisation
was	O
instrumental	O
to	O
the	O
U.S.	B-organisation
Air	I-organisation
Force	I-organisation
Missile	I-organisation
Directorate	I-organisation
at	O
Hill	B-location
Air	I-location
Force	I-location
Base	I-location
near	O
Ogden	B-location
,	O
Utah	B-location
to	O
produce	O
in	O
top	O
military	O
secrecy	O
,	O
the	O
Intelligent	B-product
Systems	I-product
Technology	I-product
Software	I-product
that	O
was	O
foundational	O
to	O
the	O
later	O
named	O
Reagan	B-product
Star	I-product
Wars	I-product
program	I-product
.	O

Over	O
the	O
decades	O
he	O
has	O
researched	O
and	O
developed	O
emerging	O
fields	O
of	O
computer	B-field
science	I-field
from	O
compiler	O
,	O
programming	O
languages	O
and	O
system	O
architecture	O
John	B-researcher
F.	I-researcher
Sowa	I-researcher
and	O
John	B-researcher
Zachman	I-researcher
(	O
1992	O
)	O
.	O

The	O
Sobel	B-algorithm
operator	I-algorithm
,	O
sometimes	O
called	O
the	O
Sobel-Feldman	B-algorithm
operator	I-algorithm
or	O
Sobel	B-algorithm
filter	I-algorithm
,	O
is	O
used	O
in	O
image	B-field
processing	I-field
and	O
computer	B-field
vision	I-field
,	O
particularly	O
within	O
edge	B-misc
detection	I-misc
algorithms	I-misc
where	O
it	O
creates	O
an	O
image	O
emphasising	O
edges	O
.	O

LDA	B-algorithm
is	O
a	O
supervised	B-field
learning	I-field
algorithm	O
that	O
utilizes	O
the	O
labels	O
of	O
the	O
data	O
,	O
while	O
PCA	B-algorithm
is	O
an	O
learning	O
algorithm	O
that	O
ignores	O
the	O
labels	O
.	O

Other	O
linear	O
classification	O
algorithms	O
include	O
Winnow	B-algorithm
,	O
support	B-algorithm
vector	I-algorithm
machine	I-algorithm
and	O
logistic	B-algorithm
regression	I-algorithm
.	O

VTK	B-product
consists	O
of	O
a	O
C	B-programlang
+	I-programlang
+	I-programlang
class	O
library	O
and	O
several	O
interpreted	O
interface	O
layers	O
including	O
Tcl	B-product
/	I-product
Tk	I-product
,	O
Java	B-programlang
,	O
and	O
Python	B-programlang
.	O

Also	O
,	O
text	O
produced	O
by	O
processing	O
spontaneous	O
speech	O
using	O
automatic	B-task
speech	I-task
recognition	I-task
and	O
printed	O
or	O
handwritten	O
text	O
using	O
optical	B-task
character	I-task
recognition	I-task
contains	O
processing	O
noise	O
.	O

Miller	B-researcher
wrote	O
several	O
books	O
and	O
directed	O
the	O
development	O
of	O
WordNet	B-product
,	O
an	O
online	O
word-linkage	O
database	O
usable	O
by	O
computer	O
programs	O
.	O

Contemporary	O
automata	B-field
are	O
represented	O
by	O
the	O
works	O
of	O
Cabaret	B-organisation
Mechanical	I-organisation
Theatre	I-organisation
in	O
the	O
United	B-country
Kingdom	I-country
,	O
Dug	B-person
North	I-person
and	O
Chomick	B-person
+	I-person
Meder	I-person
,	O
Arthur	B-person
Ganson	I-person
,	O
Joe	B-person
Jones	I-person
in	O
the	O
United	B-country
States	I-country
,	O
Le	B-location
Défenseur	I-location
du	I-location
Temps	I-location
by	O
French	B-misc
artist	O
Jacques	B-person
Monestier	I-person
,	O
and	O
François	B-person
Junod	I-person
in	O
Switzerland	B-country
.	O

MATLAB	B-product
does	O
include	O
standard	O
codefor	O
/	O
code	O
and	O
codewhile	O
/	O
code	O
loops	O
,	O
but	O
(	O
as	O
in	O
other	O
similar	O
applications	O
such	O
as	O
R	B-programlang
)	O
,	O
using	O
the	O
vectorized	O
notation	O
is	O
encouraged	O
and	O
is	O
often	O
faster	O
to	O
execute	O
.	O

Pausch	B-researcher
received	O
two	O
awards	O
from	O
Association	B-conference
for	I-conference
Computing	I-conference
Machinery	I-conference
in	O
2007	O
for	O
his	O
achievements	O
in	O
computing	B-field
education	I-field
:	O
the	O
Karl	B-misc
V.	I-misc
Karlstrom	I-misc
Outstanding	I-misc
Educator	I-misc
Award	I-misc
and	O
the	O
ACM	B-misc
SIGCSE	I-misc
Award	I-misc
for	I-misc
Outstanding	I-misc
Contributions	I-misc
to	I-misc
Computer	I-misc
Science	I-misc
Education	I-misc
.	O

In	O
1960	O
,	O
Devol	B-person
personally	O
sold	O
the	O
first	O
Unimate	B-product
robot	B-product
,	O
which	O
was	O
shipped	O
in	O
1961	O
to	O
General	B-organisation
Motors	I-organisation
.	O

Semantic	B-algorithm
networks	I-algorithm
are	O
used	O
in	O
natural	B-field
language	I-field
processing	I-field
applications	O
such	O
as	O
semantic	B-task
parsing	I-task
.	O

Some	O
successful	O
applications	O
of	O
deep	B-field
learning	I-field
are	O
computer	B-field
vision	I-field
and	O
speech	B-task
recognition	I-task
.	O
Honglak	B-researcher
Lee	I-researcher
,	O
Roger	B-researcher
Grosse	I-researcher
,	O
Rajesh	B-researcher
Ranganath	I-researcher
,	O
Andrew	B-researcher
Y.	I-researcher
Ng	I-researcher
.	O

In	O
addition	O
to	O
maintaining	O
the	O
Discovery	B-product
One	I-product
spacecraft	I-product
systems	I-product
during	O
the	O
interplanetary	O
mission	O
to	O
Jupiter	B-misc
(	O
or	O
Saturn	B-misc
in	O
the	O
novel	O
)	O
,	O
HAL	B-product
is	O
capable	O
of	O
speech	B-task
synthesis	I-task
,	O
speech	B-task
recognition	I-task
,	O
facial	B-task
recognition	I-task
,	O
natural	B-field
language	I-field
processing	I-field
,	O
lip	B-task
reading	I-task
,	O
art	B-field
appreciation	I-field
,	O
Affective	B-task
computing	I-task
,	O
automated	B-task
reasoning	I-task
,	O
spacecraft	B-task
piloting	I-task
and	O
playing	B-task
chess	I-task
.	O

Dr.	B-researcher
Julesz	I-researcher
emigrated	O
from	O
Hungary	B-country
to	O
the	B-country
United	I-country
States	I-country
following	O
the	O
1956	O
Soviet	B-country
invasion	O
.	O

Sigmoid	B-algorithm
function	I-algorithm
activation	O
functions	O
use	O
a	O
second	O
non-linearity	O
for	O
large	O
inputs	O
:	O
math	O
\	O
phi	O
(	O
v	O
_	O
i	O
)	O
=	O
(	O
1	O
+	O
\	O
exp	O
(	O
-v	O
_	O
i	O
)	O
)	O
^	O
{	O
-1	O
}	O
/	O
math	O
.	O

These	O
probabilities	O
are	O
used	O
to	O
determine	O
what	O
the	O
target	O
is	O
using	O
a	O
maximum	B-algorithm
likelihood	I-algorithm
decision	I-algorithm
.	O

In	O
1984	O
he	O
moved	O
to	O
the	O
University	B-university
of	I-university
Konstanz	I-university
and	O
in	O
1990	O
to	O
the	O
University	B-university
of	I-university
Salzburg	I-university
.	O

Some	O
popular	O
fitness	O
functions	O
based	O
on	O
the	O
confusion	B-metrics
matrix	I-metrics
include	O
sensitivity	B-metrics
/	I-metrics
specificity	I-metrics
,	O
recall	B-metrics
/	I-metrics
precision	I-metrics
,	O
F-measure	B-metrics
,	O
Jaccard	B-metrics
similarity	I-metrics
,	O
Matthews	B-metrics
correlation	I-metrics
coefficient	I-metrics
,	O
and	O
cost	B-metrics
/	I-metrics
gain	I-metrics
matrix	I-metrics
which	O
combines	O
the	O
costs	O
and	O
gains	O
assigned	O
to	O
the	O
4	O
different	O
types	O
of	O
classifications	O
.	O

Common	O
numerical	O
programming	O
environments	O
such	O
as	O
MATLAB	B-product
,	O
SciLab	B-product
,	O
NumPy	B-product
,	O
Sklearn	B-product
and	O
the	O
R	B-programlang
language	I-programlang
provide	O
some	O
of	O
the	O
simpler	O
feature	O
extraction	O
techniques	O
(	O
e.g.	O
principal	B-algorithm
component	I-algorithm
analysis	I-algorithm
)	O
via	O
built-in	O
commands	O
.	O

Industrial	B-product
robots	I-product
have	O
been	O
implemented	O
to	O
collaborate	O
with	O
humans	O
to	O
perform	O
industrial	O
manufacturing	O
tasks	O
.	O

In	O
the	O
first	O
published	O
paper	O
on	O
CGs	B-field
,	O
John	B-researcher
F.	I-researcher
Sowa	I-researcher
applied	O
them	O
to	O
a	O
wide	O
range	O
of	O
topics	O
in	O
artificial	B-field
intelligence	I-field
,	O
computer	B-field
science	I-field
,	O
and	O
cognitive	B-field
science	I-field
.	O

NIST	B-metrics
also	O
differs	O
from	O
BLEU	B-metrics
in	O
its	O
calculation	O
of	O
the	O
brevity	B-misc
penalty	I-misc
,	O
insofar	O
as	O
small	O
variations	O
in	O
translation	O
length	O
do	O
not	O
impact	O
the	O
overall	O
score	O
as	O
much	O
.	O

The	O
IJCAI	B-misc
Award	I-misc
for	I-misc
Research	I-misc
Excellence	I-misc
is	O
a	O
biannual	O
award	O
given	O
at	O
the	O
IJCAI	B-conference
conference	O
to	O
researcher	O
in	O
artificial	B-field
intelligence	I-field
as	O
a	O
recognition	O
of	O
excellence	O
of	O
their	O
career	O
.	O

Lenat	B-researcher
was	O
one	O
of	O
the	O
original	O
Fellows	O
of	O
the	O
AAAI	B-conference
,	O
and	O
is	O
the	O
only	O
individual	O
to	O
have	O
on	O
the	O
Scientific	B-organisation
Advisory	I-organisation
Boards	I-organisation
of	I-organisation
both	I-organisation
Microsoft	I-organisation
and	I-organisation
Apple	I-organisation
.	O

Autoencoders	B-algorithm
are	O
trained	O
to	O
minimise	O
reconstruction	O
errors	O
(	O
such	O
as	O
Mean	B-metrics
squared	I-metrics
error	I-metrics
)	O
,	O
often	O
referred	O
to	O
as	O
the	O
loss	B-misc
:	O

An	O
alternative	O
to	O
the	O
use	O
of	O
the	O
definitions	O
is	O
to	O
consider	O
general	O
word-sense	O
relatedness	O
and	O
to	O
compute	O
the	O
similarity	O
of	O
each	O
pair	O
of	O
word	O
senses	O
based	O
on	O
a	O
given	O
lexical	B-misc
knowledge	I-misc
base	I-misc
such	O
as	O
WordNet	B-product
.	O

TD-Lambda	B-algorithm
is	O
a	O
learning	O
algorithm	O
invented	O
by	O
Richard	B-researcher
S.	I-researcher
Sutton	I-researcher
based	O
on	O
earlier	O
work	O
on	O
temporal	O
difference	O
learning	O
by	O
Arthur	B-researcher
Samuel	I-researcher
.	O

In	O
data	B-field
mining	I-field
and	O
statistics	B-field
,	O
hierarchical	B-task
clustering	I-task
(	O
also	O
called	O
hierarchical	B-task
cluster	I-task
analysis	I-task
or	O
HCA	B-task
)	O
is	O
a	O
method	O
of	O
cluster	B-task
analysis	I-task
which	O
seeks	O
to	O
build	O
a	O
hierarchy	O
of	O
clusters	O
.	O

The	O
concept	O
of	O
deconvolution	B-algorithm
is	O
widely	O
used	O
in	O
the	O
techniques	O
of	O
signal	B-field
processing	I-field
and	O
image	B-field
processing	I-field
.	O

Cognitive	B-algorithm
maps	I-algorithm
serve	O
the	O
construction	O
and	O
accumulation	O
of	O
spatial	O
knowledge	O
,	O
allowing	O
the	O
mind	O
's	O
eye	O
to	O
visualize	O
images	O
in	O
order	O
to	O
reduce	O
cognitive	B-misc
load	I-misc
,	O
enhance	O
recall	B-metrics
and	O
learning	O
of	O
information	O
.	O

,	O
typically	O
providing	O
bindings	O
to	O
languages	O
such	O
as	O
Python	B-programlang
,	O
C	B-programlang
+	I-programlang
+	I-programlang
,	O
Java	B-programlang
)	O
.	O

A	O
voice-user	B-product
interface	I-product
(	O
VUI	B-product
)	O
makes	O
spoken	O
human	O
interaction	O
with	O
computers	O
possible	O
,	O
using	O
speech	B-task
recognition	I-task
to	O
understand	O
spoken	O
commands	O
and	O
Question	B-task
answering	I-task
,	O
and	O
typically	O
text	B-task
to	I-task
speech	I-task
to	O
play	O
a	O
reply	O
.	O

Jess	B-programlang
is	O
a	O
rule	B-misc
engine	I-misc
for	O
the	O
Java	B-programlang
platform	O
that	O
was	O
developed	O
by	O
Ernest	B-researcher
Friedman-Hill	I-researcher
of	O
Sandia	B-organisation
National	I-organisation
.	O

For	O
multilayer	B-algorithm
perceptron	I-algorithm
s	O
,	O
where	O
a	O
hidden	O
layer	O
exists	O
,	O
more	O
sophisticated	O
algorithms	O
such	O
as	O
backpropagation	B-algorithm
must	O
be	O
used	O
.	O

Google	B-product
Translate	I-product
's	O
neural	B-product
machine	I-product
translation	I-product
system	I-product
uses	O
a	O
large	O
end-to-end	B-algorithm
artificial	I-algorithm
neural	I-algorithm
network	I-algorithm
that	O
attempts	O
to	O
perform	O
deep	B-field
learning	I-field
,	O
in	O
particular	O
,	O
long	B-algorithm
short-term	I-algorithm
memory	I-algorithm
networks	I-algorithm
.	O

Various	O
methods	O
for	O
doing	O
so	O
were	O
developed	O
in	O
the	O
1980s	O
and	O
early	O
1990s	O
by	O
Werbos	B-researcher
,	O
Williams	B-researcher
,	O
Robinson	B-researcher
,	O
Jürgen	B-researcher
Schmidhuber	I-researcher
,	O
Sepp	B-researcher
Hochreiter	I-researcher
,	O
Pearlmutter	B-researcher
and	O
others	O
.	O

|	O
Apple	B-organisation
Apple	B-organisation
Inc	I-organisation
originally	O
licensed	O
software	O
from	O
Nuance	B-organisation
to	O
provide	O
speech	B-task
recognition	I-task
capability	O
to	O
its	O
digital	O
assistant	O
Siri	B-product
.	O

Columbia	B-organisation
released	O
several	O
3D	B-misc
westerns	I-misc
produced	O
by	O
Sam	B-person
Katzman	I-person
and	O
directed	O
by	O
William	B-person
Castle	I-person
.	O

It	O
incorporates	O
knowledge	O
and	O
research	O
in	O
the	O
computer	B-field
science	I-field
,	O
linguistics	B-field
and	O
computer	B-field
engineering	I-field
fields	O
.	O

Here	O
is	O
an	O
example	O
of	O
R	B-programlang
code	O
:	O

The	O
ROC	B-metrics
curve	I-metrics
is	O
created	O
by	O
plotting	O
the	O
TRUE	B-metrics
positive	I-metrics
rate	I-metrics
(	O
TPR	B-metrics
)	O
against	O
the	O
FALSE	B-metrics
positive	I-metrics
rate	I-metrics
(	O
FPR	B-metrics
)	O
at	O
various	O
threshold	O
settings	O
.	O

Research	O
stagnated	O
after	O
machine	B-field
learning	I-field
research	O
by	O
Marvin	B-researcher
Minsky	I-researcher
and	O
Seymour	B-researcher
Papert	I-researcher
(	O
1969	O
)	O
,	O

Other	O
programming	O
environments	O
that	O
are	O
used	O
to	O
build	O
DAQ	B-task
applications	O
include	O
ladder	B-programlang
logic	I-programlang
,	O
Visual	B-product
C	I-product
+	I-product
+	I-product
,	O
Visual	B-programlang
Basic	I-programlang
,	O
LabVIEW	B-product
,	O
and	O
MATLAB	B-product
.	O

The	O
metric	O
was	O
designed	O
to	O
fix	O
some	O
of	O
the	O
problems	O
found	O
in	O
the	O
more	O
popular	O
BLEU	B-metrics
metric	I-metrics
,	O
and	O
also	O
produce	O
good	O
correlation	O
with	O
human	O
judgement	O
at	O
the	O
sentence	O
or	O
segment	O
level	O
.	O

Techniques	O
such	O
as	O
dynamic	B-algorithm
Markov	I-algorithm
Networks	I-algorithm
,	O
Convolutional	B-algorithm
neural	I-algorithm
network	I-algorithm
and	O
Long	B-algorithm
short-term	I-algorithm
memory	I-algorithm
are	O
often	O
employed	O
to	O
exploit	O
the	O
semantic	O
correlations	O
between	O
consecutive	O
video	O
frames	O
.	O

Mass-produced	O
printed	B-product
circuit	I-product
board	I-product
s	O
(	O
PCBs	B-product
)	O
are	O
almost	O
exclusively	O
manufactured	O
by	O
pick-and-place	B-product
robots	I-product
,	O
typically	O
with	O
SCARA	B-product
manipulators	O
,	O
which	O
remove	O
tiny	O
electronic	O
component	O
s	O
from	O
strips	O
or	O
trays	O
,	O
and	O
place	O
them	O
on	O
to	O
PCBs	B-product
with	O
great	O
accuracy	O
.	O

In	O
the	O
context	O
of	O
machine	B-field
learning	I-field
,	O
where	O
it	O
is	O
most	O
widely	O
applied	O
today	O
,	O
LDA	B-algorithm
was	O
rediscovered	O
independently	O
by	O
David	B-researcher
Blei	I-researcher
,	O
Andrew	B-researcher
Ng	I-researcher
and	O
Michael	B-researcher
I.	I-researcher
Jordan	I-researcher
in	O
2003	O
,	O
and	O
presented	O
as	O
a	O
graphical	B-algorithm
model	I-algorithm
for	O
topic	B-task
discovery	I-task
.	O

The	O
measured	O
performance	O
on	O
test	O
data	O
of	O
eight	O
naive	O
WSI	B-task
across	O
various	O
tauopathies	B-misc
resulted	O
in	O
the	O
recall	B-metrics
,	O
precision	B-metrics
,	O
and	O
an	O
F1	B-metrics
score	I-metrics
of	O
0.92	O
,	O
0.72	O
,	O
and	O
0.81	O
,	O
respectively	O
.	O

With	O
the	O
help	O
of	O
advanced	O
AR	B-field
technologies	O
(	O
e.g.	O
adding	O
computer	B-field
vision	I-field
,	O
incorporating	O
AR	B-field
cameras	O
into	O
smartphone	O
and	O
object	B-task
recognition	I-task
)	O
the	O
information	O
about	O
the	O
surrounding	O
real	O
world	O
of	O
the	O
user	O
becomes	O
interactive	O
and	O
digitally	O
manipulated	O
.	O

In	O
2014	O
,	O
Schmidhuber	B-researcher
formed	O
a	O
company	O
,	O
Nnaisense	B-organisation
,	O
to	O
work	O
on	O
commercial	O
applications	O
of	O
artificial	B-field
intelligence	I-field
in	O
fields	O
such	O
as	O
finance	O
,	O
heavy	O
industry	O
and	O
self-driving	B-product
car	I-product
s	O
.	O

Not	O
only	O
does	O
this	O
alter	O
the	O
performance	O
of	O
all	O
subsequent	O
tests	O
on	O
the	O
retained	O
explanatory	O
model	O
,	O
it	O
may	O
introduce	O
bias	O
and	O
alter	O
mean	B-metrics
square	I-metrics
error	I-metrics
in	O
estimation	O
.	O

Bigrams	B-misc
are	O
used	O
in	O
most	O
successful	O
language	B-algorithm
model	I-algorithm
s	O
for	O
speech	B-task
recognition	I-task
.	O

His	O
research	O
in	O
cognitive	B-field
psychology	I-field
has	O
won	O
the	O
Early	B-misc
Career	I-misc
Award	I-misc
(	O
1984	O
)	O
and	O
Boyd	B-misc
McCandless	I-misc
Award	I-misc
1986	O
)	O
from	O
the	O
American	B-organisation
Psychological	I-organisation
Association	I-organisation
,	O
the	O
Troland	B-misc
Research	I-misc
Award	I-misc
(	O
1993	O
)	O
from	O
the	O
National	B-organisation
Academy	I-organisation
of	I-organisation
Sciences	I-organisation
,	O
the	O
Henry	B-misc
Dale	I-misc
Prize	I-misc
(	O
2004	O
)	O
from	O
the	O
Royal	B-organisation
Institution	I-organisation
of	I-organisation
Great	I-organisation
Britain	I-organisation
,	O
and	O
the	O
George	B-misc
Miller	I-misc
Prize	I-misc
(	O
2010	O
)	O
from	O
the	O
Cognitive	B-organisation
Neuroscience	I-organisation
Society	I-organisation
.	O

An	O
eigenface	B-misc
(	O
The	O
approach	O
of	O
using	O
eigenfaces	B-misc
for	O
Facial	B-product
recognition	I-product
system	I-product
was	O
developed	O
by	O
Sirovich	B-researcher
and	O
Kirby	B-researcher
(	O
1987	O
)	O
and	O
used	O
by	O
Matthew	B-researcher
Turk	I-researcher
and	O
Alex	B-researcher
Pentland	I-researcher
in	O
face	B-task
classification	I-task
.	O
Turk	B-researcher
,	I-researcher
Matthew	I-researcher
A	I-researcher
and	O
Pentland	B-researcher
,	I-researcher
Alex	I-researcher
P.	I-researcher
Face	B-task
recognition	I-task
using	O
eigenfaces	B-misc
.	O

A	O
lexical	O
dictionary	O
such	O
as	O
WordNet	B-product
can	O
then	O
be	O
used	O
for	O
understanding	O
the	O
context	O
.	O

Hyponymy	B-misc
is	O
the	O
most	O
frequently	O
encoded	O
relation	O
among	O
synsets	B-misc
used	O
in	O
lexical	O
databases	O
such	O
as	O
WordNet	B-product
.	O

OPeNDAP	B-organisation
offers	O
open-source	O
libraries	O
in	O
C	B-programlang
+	I-programlang
+	I-programlang
and	O
Java	B-programlang
,	O
but	O
many	O
clients	O
rely	O
on	O
community	O
developed	O
libraries	O
such	O
as	O
libraries	O
include	O
embedded	O
capabilities	O
for	O
retrieving	O
(	O
array-style	O
)	O
data	O
from	O
DAP	B-misc
servers	O
.	O

In	O
that	O
page	O
,	O
Samurai	B-misc
Damashii	I-misc
exaggerated	O
the	O
Senkousha	B-product
as	O
the	O
crystallization	O
of	O
China	B-country
's	O
four	O
thousand	O
years	O
of	O
scientific	O
knowledge	O
,	O
commented	O
on	O
the	O
crude	O
design	O
(	O
e.g.	O
the	O
Chinese	B-misc
Cannon	I-misc
on	O
its	O
crotch	O
)	O
,	O
and	O
put	O
its	O
image	O
among	O
images	O
of	O
Honda	B-organisation
'	O
s	O
ASIMO	B-product
and	O
Sony	B-organisation
'	O
s	O
QRIO	B-product
SDR-3X	I-product
for	O
juxtaposition	O
.	O

There	O
are	O
also	O
many	O
programming	O
libraries	O
that	O
contain	O
neural	B-algorithm
network	I-algorithm
functionality	O
and	O
that	O
can	O
be	O
used	O
in	O
custom	O
implementations	O
(	O
such	O
as	O
TensorFlow	B-product
,	O
Theano	B-product
,	O
etc	O
.	O

He	O
is	O
a	O
Fellow	O
of	O
the	O
Association	B-conference
for	I-conference
Computing	I-conference
Machinery	I-conference
,	O
IEEE	B-organisation
,	O
American	B-conference
Association	I-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Science	I-conference
,	O
IAPR	B-conference
and	O
SPIE	B-conference
.	O

A	O
trial	O
by	O
RET	B-organisation
in	O
2011	O
with	O
Facial	B-product
recognition	I-product
system	I-product
cameras	O
mounted	O
on	O
trams	O
made	O
sure	O
that	O
people	O
were	O
banned	O
from	O
the	O
city	O
trams	O
did	O
not	O
sneak	O
on	O
anyway	O
.	O

The	O
film	O
,	O
adapted	O
from	O
the	O
popular	O
Cole	B-person
Porter	I-person
Broadway	B-organisation
musical	O
,	O
starred	O
the	O
MGM	O
songbird	O
team	O
of	O
Howard	B-person
Keel	I-person
and	O
Kathryn	B-person
Grayson	I-person
as	O
the	O
leads	O
,	O
supported	O
by	O
Ann	B-person
Miller	I-person
,	O
Keenan	B-person
Wynn	I-person
,	O
Bobby	B-person
Van	I-person
,	O
James	B-person
Whitmore	I-person
,	O
Kurt	B-person
Kasznar	I-person
and	O
Tommy	B-person
Rall	I-person
.	O

Such	O
applications	O
should	O
streamline	O
the	O
call	O
flows	O
,	O
minimize	O
prompts	O
,	O
eliminate	O
unnecessary	O
iterations	O
and	O
allow	O
elaborate	O
mixed	B-product
initiative	I-product
dialog	I-product
system	I-product
,	O
which	O
enable	O
callers	O
to	O
enter	O
several	O
pieces	O
of	O
information	O
in	O
a	O
single	O
utterance	O
and	O
in	O
any	O
order	O
or	O
combination	O
.	O

As	O
such	O
,	O
traditional	O
gradient	B-algorithm
descent	I-algorithm
(	O
or	O
Stochastic	B-algorithm
gradient	I-algorithm
descent	I-algorithm
)	O
methods	O
can	O
be	O
adapted	O
,	O
where	O
of	O
taking	O
a	O
step	O
in	O
the	O
direction	O
of	O
the	O
function	O
's	O
gradient	O
,	O
a	O
step	O
is	O
taken	O
in	O
the	O
direction	O
of	O
a	O
vector	O
selected	O
from	O
the	O
function	O
's	O
sub-gradient	O
.	O

If	O
it	O
is	O
assumed	O
that	O
distortion	O
is	O
measured	O
by	O
mean	B-metrics
squared	I-metrics
error	I-metrics
,	O
the	O
distortion	B-misc
D	I-misc
,	O
is	O
given	O
by	O
:	O

MLPs	B-algorithm
were	O
a	O
popular	O
machine	B-field
learning	I-field
solution	O
in	O
the	O
1980s	O
,	O
finding	O
applications	O
in	O
diverse	O
fields	O
such	O
as	O
speech	B-task
recognition	I-task
,	O
image	B-task
recognition	I-task
,	O
and	O
machine	B-task
translation	I-task
software	O
,	O
Neural	B-product
networks	I-product
.	O

Allen	B-researcher
received	O
his	O
Ph.D.	O
from	O
the	O
University	B-university
of	I-university
Toronto	I-university
in	O
1979	O
,	O
under	O
the	O
supervision	O
of	O
C.	B-researcher
Raymond	I-researcher
Perrault	I-researcher
,	O

OpenCV	B-product
supports	O
some	O
models	O
from	O
deep	B-field
learning	I-field
frameworks	O
like	O
TensorFlow	B-product
,	O
Torch	B-product
,	O
PyTorch	B-product
(	O
after	O
converting	O
to	O
an	O
ONNX	B-product
model	O
)	O
and	O
Caffe	B-product
according	O
to	O
a	O
defined	O
list	O
of	O
supported	O
layers	O
.	O

Previously	O
,	O
Christensen	B-researcher
was	O
the	O
Founding	O
Chairman	O
of	O
European	B-organisation
Robotics	I-organisation
Research	I-organisation
Network	I-organisation
(	O
EURON	B-organisation
)	O
and	O
an	O
IEEE	B-organisation
Robotics	I-organisation
and	I-organisation
Automation	I-organisation
Society	I-organisation
Distinguished	O
Lecturer	O
in	O
Robotics	B-field
.	O

He	O
received	O
his	O
master	O
's	O
degree	O
in	O
mathematics	B-field
from	O
the	O
Samarkand	B-university
State	I-university
University	I-university
,	O
Samarkand	B-location
,	O
Uzbek	B-country
Soviet	I-country
Socialist	I-country
Republic	I-country
in	O
1958	O
and	O
Ph.D	B-misc
in	O
statistics	B-field
at	O
the	O
Institute	B-organisation
of	I-organisation
Control	I-organisation
Sciences	I-organisation
,	O
Moscow	B-location
in	O
1964	O
.	O

Increasingly	O
,	O
however	O
,	O
work	O
at	O
Cycorp	B-organisation
involves	O
giving	O
the	O
Cyc	B-product
system	I-product
the	O
ability	O
to	O
communicate	O
with	O
end	O
users	O
in	O
natural	O
language	O
,	O
and	O
to	O
assist	O
with	O
the	O
ongoing	O
knowledge	O
formation	O
process	O
via	O
machine	B-field
learning	I-field
and	O
natural	B-task
language	I-task
understanding	I-task
.	O

For	O
example	O
,	O
if	O
the	O
most	O
suitable	O
classifier	O
for	O
the	O
problem	O
is	O
sought	O
,	O
the	O
training	O
dataset	O
is	O
used	O
to	O
train	O
the	O
candidate	O
algorithms	O
,	O
the	O
validation	O
dataset	O
is	O
used	O
to	O
compare	O
their	O
performances	O
and	O
decide	O
which	O
one	O
to	O
take	O
and	O
,	O
finally	O
,	O
the	O
test	O
dataset	O
is	O
used	O
to	O
obtain	O
the	O
performance	O
characteristics	O
such	O
as	O
accuracy	B-metrics
,	O
sensitivity	B-metrics
,	O
specificity	B-metrics
,	O
F-measure	B-metrics
,	O
and	O
so	O
on	O
.	O

The	O
Mean	B-metrics
squared	I-metrics
error	I-metrics
is	O
0.15	O
.	O

In	O
1979	O
a	O
Micromouse	B-misc
competition	I-misc
was	O
organized	O
by	O
the	O
IEEE	B-organisation
as	O
shown	O
in	O
the	O
Spectrum	B-misc
magazine	O
.	O

The	O
Gabor	B-algorithm
space	I-algorithm
is	O
very	O
useful	O
in	O
image	B-field
processing	I-field
applications	O
such	O
as	O
optical	B-task
character	I-task
recognition	I-task
,	O
iris	B-task
recognition	I-task
and	O
fingerprint	B-task
recognition	I-task
.	O

or	O
via	O
high-level	O
interfaces	O
to	O
Java	B-programlang
and	O
Tcl	B-programlang
.	O

In	O
recent	O
research	O
,	O
kernel-based	O
methods	O
such	O
as	O
support	B-algorithm
vector	I-algorithm
machine	I-algorithm
s	O
have	O
shown	O
superior	O
performance	O
in	O
supervised	B-field
.	O

To	O
illustrate	O
the	O
basic	O
principles	O
of	O
bagging	O
,	O
below	O
is	O
an	O
analysis	O
on	O
the	O
relationship	O
between	O
ozone	B-misc
and	O
temperature	O
(	O
data	O
from	O
Rousseeuw	B-researcher
and	O
Leroy	B-researcher
(	O
1986	O
)	O
,	O
analysis	O
done	O
in	O
R	B-programlang
)	O
.	O

Denso	B-organisation
Wave	I-organisation
is	O
a	O
subsidiary	O
that	O
produces	O
automatic	O
identification	O
products	O
(	O
bar-code	B-product
reader	I-product
s	O
and	O
related	O
products	O
)	O
,	O
industrial	B-product
robot	I-product
s	O
and	O
programmable	B-product
logic	I-product
controller	I-product
s	O
.	O

Where	O
Bilingual	B-metrics
evaluation	I-metrics
understudy	I-metrics
simply	O
calculates	O
n-gram	B-metrics
precision	I-metrics
adding	O
equal	O
weight	O
to	O
each	O
one	O
,	O
NIST	B-metrics
also	O
calculates	O
how	O
informative	O
a	O
particular	O
n-gram	B-misc
is	O
.	O

In	O
particular	O
,	O
they	O
are	O
used	O
during	O
the	O
calculation	O
of	O
likelihood	O
of	O
a	O
tree	O
(	O
in	O
Bayesian	B-algorithm
and	O
maximum	B-algorithm
likelihood	I-algorithm
approaches	O
to	O
tree	O
estimation	O
)	O
and	O
they	O
are	O
used	O
to	O
estimate	O
the	O
evolutionary	O
distance	O
between	O
sequences	O
from	O
the	O
observed	O
differences	O
between	O
the	O
sequences	O
.	O

The	O
Audio	B-conference
Engineering	I-conference
Society	I-conference
recommends	O
48	O
kHz	O
sampling	O
rate	O
for	O
most	O
applications	O
but	O
gives	O
recognition	O
to	O
44.1	O
kHz	O
for	O
Compact	B-misc
Disc	I-misc
(	O
CD	B-misc
)	O
and	O
other	O
consumer	O
uses	O
,	O
32	O
kHz	O
for	O
transmission-related	O
applications	O
,	O
and	O
96	O
kHz	O
for	O
higher	O
bandwidth	O
or	O
relaxed	O
anti-aliasing	B-misc
filter	I-misc
ing	O
.	O

Resources	O
for	O
affectivity	O
of	O
words	O
and	O
concepts	O
have	O
been	O
made	O
for	O
WordNet	B-product
{	O
{	O
cite	O
journal	O

In	O
red-green	B-misc
anaglyph	I-misc
,	O
the	O
audience	O
was	O
presented	O
three	O
reels	O
of	O
tests	O
,	O
which	O
included	O
rural	O
scenes	O
,	O
test	O
shots	O
of	O
Marie	B-person
Doro	I-person
,	O
a	O
segment	O
of	O
John	B-person
B.	I-person
Mason	I-person
playing	O
a	O
number	O
of	O
passages	O
from	O
Jim	B-person
the	I-person
Penman	I-person
(	O
a	O
film	O
released	O
by	O
Famous	B-organisation
Players-Lasky	I-organisation
that	O
year	O
,	O
but	O
not	O
in	O
3D	O
)	O
,	O
Oriental	O
dancers	O
,	O
and	O
a	O
reel	O
of	O
footage	O
of	O
Niagara	B-location
Falls	I-location
.	O

This	O
is	O
a	O
particular	O
way	O
of	O
implementing	O
maximum	B-metrics
likelihood	I-metrics
estimation	I-metrics
for	O
this	O
problem	O
.	O

Crawler-friendly	B-product
Web	I-product
Servers	I-product
,	O
and	O
it	O
integrates	O
the	O
features	O
of	O
sitemaps	O
and	O
RSS	B-misc
feeds	O
into	O
a	O
decentralized	O
mechanism	O
for	O
computational	O
biologists	O
and	O
bio-informaticians	O
to	O
openly	O
broadcast	O
and	O
retrieve	O
meta-data	O
about	O
biomedical	O
resources	O
.	O

It	O
is	O
covered	O
by	O
American	B-misc
National	I-misc
Standards	I-misc
Institute	I-misc
/	I-misc
NISO	I-misc
standard	I-misc
Z39.50	I-misc
,	O
and	O
International	B-misc
Organization	I-misc
for	I-misc
Standardization	I-misc
standard	I-misc
23950	I-misc
.	O

The	O
encoder	O
and	O
decoder	O
are	O
trained	O
to	O
take	O
a	O
phrase	O
and	O
reproduce	O
the	O
one-hot	B-misc
distribution	I-misc
of	O
a	O
corresponding	O
paraphrase	O
by	O
minimizing	O
perplexity	B-metrics
using	O
simple	O
stochastic	B-algorithm
gradient	I-algorithm
descent	I-algorithm
.	O

Other	O
typical	O
applications	O
of	O
pattern	B-field
recognition	I-field
techniques	O
are	O
automatic	B-task
speech	I-task
recognition	I-task
,	O
classification	B-task
of	I-task
text	I-task
into	I-task
several	I-task
categories	I-task
(	O
e.g.	O
,	O
spam	O
/	O
non-spam	O
email	O
messages	O
)	O
,	O
the	O
handwriting	B-task
recognition	I-task
on	I-task
postal	I-task
envelopes	I-task
,	O
automatic	B-task
recognition	I-task
of	I-task
images	I-task
of	I-task
human	I-task
faces	I-task
,	O
or	O
handwriting	B-task
image	I-task
extraction	I-task
from	I-task
medical	I-task
forms	I-task
.	O

Artificial	B-algorithm
neural	I-algorithm
networks	I-algorithm
have	O
been	O
used	O
on	O
a	O
variety	O
of	O
tasks	O
,	O
including	O
computer	B-field
vision	I-field
,	O
speech	B-task
recognition	I-task
,	O
machine	B-task
translation	I-task
,	O
social	B-task
network	I-task
filtering	I-task
,	O
playing	B-task
board	I-task
and	I-task
video	I-task
games	I-task
and	O
medical	B-task
diagnosis	I-task
.	O

Examples	O
include	O
Salford	B-organisation
Systems	I-organisation
CART	B-product
(	O
which	O
licensed	O
the	O
proprietary	O
code	O
of	O
the	O
original	O
CART	B-product
authors	O
)	O
,	O
IBM	B-organisation
SPSS	B-product
Modeler	I-product
,	O
RapidMiner	B-product
,	O
SAS	B-product
Enterprise	I-product
Miner	I-product
,	O
Matlab	B-product
,	O
R	B-programlang
(	O
an	O
open-source	O
software	O
environment	O
for	O
statistical	B-field
computing	I-field
,	O
which	O
includes	O
several	O
CART	B-product
implementations	O
such	O
as	O
rpart	B-algorithm
,	O
party	B-algorithm
and	O
randomForest	B-algorithm
packages	O
)	O
,	O
Weka	B-product
(	O
a	O
free	O
and	O
open-source	O
data-mining	B-task
suite	O
,	O
contains	O
many	O
decision	B-algorithm
tree	I-algorithm
algorithms	O
)	O
,	O
Orange	B-product
,	O
KNIME	B-product
,	O
Microsoft	B-product
SQL	I-product
Server	I-product
programming	O
language	O
)	O
.	O

Linear	B-algorithm
predictive	I-algorithm
coding	I-algorithm
(	O
LPC	B-algorithm
)	O
was	O
first	O
developed	O
by	O
Fumitada	B-researcher
Itakura	I-researcher
of	O
Nagoya	B-university
University	I-university
and	O
Shuzo	B-researcher
Saito	I-researcher
of	O
Nippon	B-organisation
Telegraph	I-organisation
and	I-organisation
Telephone	I-organisation
(	O
NTT	B-organisation
)	O
in	O
1966	O
,	O
and	O
then	O
further	O
developed	O
by	O
Bishnu	B-researcher
S.	I-researcher
Atal	I-researcher
and	O
Manfred	B-researcher
R.	I-researcher
Schroeder	I-researcher
at	O
Bell	B-organisation
Labs	I-organisation
during	O
the	O
early-to-mid-1970s	O
,	O
becoming	O
a	O
basis	O
for	O
the	O
first	O
speech	B-product
synthesizer	I-product
DSP	I-product
chips	I-product
in	O
the	O
late	O
1970s	O
.	O

An	O
F-score	B-metrics
is	O
a	O
combination	O
of	O
the	O
precision	B-metrics
and	O
the	O
recall	B-metrics
,	O
providing	O
a	O
single	O
score	O
.	O

Image	B-field
analysis	I-field
tasks	O
can	O
be	O
as	O
simple	O
as	O
reading	B-task
bar	I-task
code	I-task
d	I-task
tags	I-task
or	O
as	O
sophisticated	O
as	O
facial	B-product
recognition	I-product
system	I-product
.	O

The	O
special	O
case	O
of	O
linear	O
support-vector	B-algorithm
machines	I-algorithm
can	O
be	O
solved	O
more	O
efficiently	O
by	O
the	O
same	O
kind	O
of	O
algorithms	O
to	O
optimize	O
its	O
close	O
cousin	O
,	O
logistic	B-algorithm
regression	I-algorithm
;	O
this	O
class	O
of	O
algorithms	O
includes	O
Stochastic	B-algorithm
gradient	I-algorithm
descent	I-algorithm
(	O
e.g.	O
,	O
PEGASOS	B-algorithm
)	O
.	O

When	O
Siri	B-product
on	O
an	O
iOS	B-product
device	O
is	O
asked	O
Do	O
you	O
have	O
a	O
pet	O
?	O
,	O
one	O
the	O
responses	O
is	O
I	O
used	O
to	O
have	O
an	O
AIBO	B-product
.	O

In	O
information	B-task
retrieval	I-task
,	O
the	O
positive	B-metrics
predictive	I-metrics
value	I-metrics
is	O
called	O
precision	B-metrics
,	O
and	O
sensitivity	B-metrics
is	O
called	O
recall	B-metrics
.	O

In	O
particular	O
,	O
his	O
research	O
focused	O
on	O
areas	O
such	O
as	O
text	B-field
mining	I-field
(	O
extraction	B-task
,	O
categorization	B-task
,	O
novelty	B-task
detection	I-task
)	O
and	O
in	O
new	O
theoretical	O
frameworks	O
such	O
as	O
a	O
unified	O
utility-based	O
theory	O
bridging	O
information	B-task
retrieval	I-task
,	O
Automatic	B-task
summarization	I-task
,	O
free-text	B-task
Question	I-task
Answering	I-task
and	O
related	O
tasks	O
.	O

Delta	B-product
robot	I-product
s	O
have	O
base-mounted	O
rotary	B-product
actuator	I-product
s	O
that	O
move	O
a	O
light	O
,	O
stiff	O
,	O
parallelogram	B-misc
arm	I-misc
.	O

The	O
four	O
outcomes	O
can	O
be	O
formulated	O
in	O
a	O
2	B-metrics
×	I-metrics
2	I-metrics
contingency	I-metrics
table	I-metrics
or	O
confusion	B-metrics
matrix	I-metrics
,	O
as	O
follows	O
:	O

The	O
actual	O
data	B-field
mining	I-field
task	O
is	O
the	O
semi-automatic	O
or	O
automatic	O
analysis	O
of	O
large	O
quantities	O
of	O
data	O
to	O
extract	O
unknown	O
,	O
interesting	O
patterns	O
such	O
as	O
groups	O
of	O
data	O
records	O
(	O
cluster	B-task
analysis	I-task
)	O
,	O
unusual	O
records	O
(	O
anomaly	B-task
detection	I-task
)	O
,	O
and	O
dependencies	O
(	O
association	B-task
rule	I-task
mining	I-task
,	O
sequential	B-task
pattern	I-task
mining	I-task
)	O
.	O

For	O
a	O
recommender	B-product
system	I-product
,	O
sentiment	B-task
analysis	I-task
has	O
been	O
proven	O
to	O
be	O
a	O
valuable	O
technique	O
.	O

By	O
chance	O
,	O
the	O
Germans	B-misc
had	O
chosen	O
the	O
operating	O
frequency	O
of	O
the	O
Wotan	B-product
system	O
very	O
badly	O
;	O
it	O
operated	O
on	O
45	O
MHz	O
,	O
which	O
just	O
happened	O
to	O
be	O
the	O
frequency	O
of	O
the	O
powerful-but-dormant	O
BBC	B-organisation
television	O
transmitter	O
at	O
Alexandra	B-location
Palace	I-location
.	O

The	O
four	O
outcomes	O
can	O
be	O
formulated	O
in	O
a	O
2	B-metrics
×	I-metrics
2	I-metrics
contingency	I-metrics
table	I-metrics
or	O
confusion	B-metrics
matrix	I-metrics
,	O
as	O
follows	O
:	O

In	O
Semantic	B-misc
Web	I-misc
applications	I-misc
,	O
and	O
in	O
relatively	O
popular	O
applications	O
of	O
RDF	B-misc
like	O
RSS	B-product
and	O
FOAF	B-product
(	O
Friend	B-product
a	I-product
Friend	I-product
)	O
,	O
resources	O
tend	O
to	O
be	O
represented	O
by	O
URIs	B-misc
that	O
intentionally	O
denote	O
,	O
and	O
can	O
be	O
used	O
to	O
access	O
,	O
actual	O
data	O
on	O
the	O
World	B-product
Wide	I-product
Web	I-product
.	O

The	O
Association	B-conference
for	I-conference
the	I-conference
Advancement	I-conference
of	I-conference
Artificial	I-conference
Intelligence	I-conference
has	O
studied	O
this	O
topic	O
in	O
depth	O

Starting	O
as	O
a	O
curiosity	O
,	O
the	O
speech	B-product
system	I-product
of	I-product
Apple	I-product
Macintosh	I-product
has	O
evolved	O
into	O
a	O
fully	O
supported	O
program	O
PlainTalk	B-product
,	O
for	O
people	O
with	O
vision	O
problems	O
.	O

Other	O
areas	O
of	O
usage	O
for	O
ontologies	O
within	O
NLP	B-field
include	O
information	B-task
retrieval	I-task
,	O
information	B-task
extraction	I-task
and	O
automatic	B-task
summarization	I-task
.	O

The	O
Institute	O
has	O
collaborated	O
closely	O
with	O
the	O
Janelia	B-organisation
Farm	I-organisation
Campus	I-organisation
of	I-organisation
Howard	I-organisation
Hughes	I-organisation
Medical	I-organisation
Institute	I-organisation
,	O
the	O
Allen	B-organisation
Institute	I-organisation
for	I-organisation
Brain	I-organisation
Science	I-organisation
and	O
the	O
National	B-organisation
Institutes	I-organisation
of	I-organisation
Health	I-organisation
to	O
develop	O
better	O
methods	O
of	O
reconstructing	O
neuronal	O
architectures	O
.	O

Recently	O
,	O
Google	B-organisation
announced	O
that	O
Google	B-product
Translate	I-product
translates	O
roughly	O
enough	O
text	O
to	O
fill	O
1	O
million	O
books	O
in	O
one	O
day	O
(	O
2012	O
)	O
.	O

Events	O
are	O
held	O
worldwide	O
,	O
and	O
are	O
most	O
popular	O
in	O
the	O
United	B-country
Kingdom	I-country
,	O
United	B-country
States	I-country
,	O
Japan	B-country
,	O
Singapore	B-country
,	O
India	B-country
,	O
South	B-country
Korea	I-country
and	O
becoming	O
popular	O
in	O
subcontinent	O
countries	O
such	O
as	O
Sri	B-country
Lanka	I-country
.	O

These	O
packages	O
are	O
developed	O
primarily	O
in	O
R	B-programlang
,	O
and	O
sometimes	O
in	O
Java	B-programlang
,	O
C	B-programlang
,	O
C	B-programlang
+	I-programlang
+	I-programlang
,	O
and	O
Fortran	B-programlang
.	O

As	O
part	O
of	O
the	O
2006	B-conference
European	I-conference
Conference	I-conference
on	I-conference
Computer	I-conference
Vision	I-conference
(	O
ECCV	B-conference
)	O
,	O
Dalal	B-researcher
and	O
Triggs	B-researcher
teamed	O
up	O
with	O
Cordelia	B-researcher
Schmid	I-researcher
to	O
apply	O
HOG	B-algorithm
detectors	I-algorithm
to	O
the	O
problem	O
of	O
human	B-task
detection	I-task
in	I-task
films	I-task
and	I-task
videos	I-task
.	O

In	O
addition	O
to	O
sensitivity	B-metrics
and	O
specificity	B-metrics
,	O
the	O
performance	O
of	O
a	O
binary	B-task
classification	I-task
test	O
can	O
be	O
measured	O
with	O
positive	B-metrics
predictive	I-metrics
value	I-metrics
(	O
PPV	B-metrics
)	O
,	O
also	O
known	O
as	O
precision	B-metrics
,	O
and	O
negative	B-metrics
predictive	I-metrics
value	I-metrics
(	O
NPV	B-metrics
)	O
.	O

Such	O
models	O
may	O
given	O
partial	O
credit	O
for	O
overlapping	O
matches	O
(	O
such	O
as	O
using	O
the	O
Jaccard	B-metrics
index	I-metrics
criterion	I-metrics
.	O

Further	O
,	O
in	O
the	O
case	O
of	O
estimation	O
based	O
on	O
a	O
single	O
sample	O
,	O
it	O
demonstrates	O
philosophical	O
issues	O
and	O
possible	O
misunderstandings	O
in	O
the	O
use	O
of	O
maximum	B-metrics
likelihood	I-metrics
estimators	I-metrics
and	I-metrics
likelihood	I-metrics
functions	I-metrics
.	O

