Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications

Rumen Dangovski, Li Jing, Preslav Nakov, Mićo Tatalović, Marin Soljačić


Abstract
Stacking long short-term memory (LSTM) cells or gated recurrent units (GRUs) as part of a recurrent neural network (RNN) has become a standard approach to solving a number of tasks ranging from language modeling to text summarization. Although LSTMs and GRUs were designed to model long-range dependencies more accurately than conventional RNNs, they nevertheless have problems copying or recalling information from the long distant past. Here, we derive a phase-coded representation of the memory state, Rotational Unit of Memory (RUM), that unifies the concepts of unitary learning and associative memory. We show experimentally that RNNs based on RUMs can solve basic sequential tasks such as memory copying and memory recall much better than LSTMs/GRUs. We further demonstrate that by replacing LSTM/GRU with RUM units we can apply neural networks to real-world problems such as language modeling and text summarization, yielding results comparable to the state of the art.
Anthology ID:
Q19-1008
Volume:
Transactions of the Association for Computational Linguistics, Volume 7
Month:
Year:
2019
Address:
Cambridge, MA
Venue:
TACL
SIG:
Publisher:
MIT Press
Note:
Pages:
121–138
Language:
URL:
https://aclanthology.org/Q19-1008
DOI:
10.1162/tacl_a_00258
Bibkey:
Cite (ACL):
Rumen Dangovski, Li Jing, Preslav Nakov, Mićo Tatalović, and Marin Soljačić. 2019. Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications. Transactions of the Association for Computational Linguistics, 7:121–138.
Cite (Informal):
Rotational Unit of Memory: A Novel Representation Unit for RNNs with Scalable Applications (Dangovski et al., TACL 2019)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/Q19-1008.pdf
Data
CNN/Daily MailPenn Treebank