Predefined Sparseness in Recurrent Sequence Models
Thomas Demeester, Johannes Deleu, Fréderic Godin, Chris Develder
Abstract
Inducing sparseness while training neural networks has been shown to yield models with a lower memory footprint but similar effectiveness to dense models. However, sparseness is typically induced starting from a dense model, and thus this advantage does not hold during training. We propose techniques to enforce sparseness upfront in recurrent sequence models for NLP applications, to also benefit training. First, in language modeling, we show how to increase hidden state sizes in recurrent layers without increasing the number of parameters, leading to more expressive models. Second, for sequence labeling, we show that word embeddings with predefined sparseness lead to similar performance as dense embeddings, at a fraction of the number of trainable parameters.- Anthology ID:
- K18-1032
- Volume:
- Proceedings of the 22nd Conference on Computational Natural Language Learning
- Month:
- October
- Year:
- 2018
- Address:
- Brussels, Belgium
- Editors:
- Anna Korhonen, Ivan Titov
- Venue:
- CoNLL
- SIG:
- SIGNLL
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 324–333
- Language:
- URL:
- https://aclanthology.org/K18-1032
- DOI:
- 10.18653/v1/K18-1032
- Cite (ACL):
- Thomas Demeester, Johannes Deleu, Fréderic Godin, and Chris Develder. 2018. Predefined Sparseness in Recurrent Sequence Models. In Proceedings of the 22nd Conference on Computational Natural Language Learning, pages 324–333, Brussels, Belgium. Association for Computational Linguistics.
- Cite (Informal):
- Predefined Sparseness in Recurrent Sequence Models (Demeester et al., CoNLL 2018)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-2/K18-1032.pdf
- Code
- tdmeeste/SparseSeqModels
- Data
- Penn Treebank