Compression of recurrent neural networks for efficient language modeling

• Published in: Applied soft computing 2019-06, Vol.79, p.354-362
• Main Authors: Grachev, Artem M., Ignatov, Dmitry I., Savchenko, Andrey V.
• Format: Article
• Language: English
• Subjects: Language modeling; Low-rank factorization; Mobile devices; Recurrent neural network compression
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2019.03.057

Recurrent neural networks have proved to be an effective method for statistical language modeling. However, in practice their memory and run-time complexity are usually too large to be implemented in real-time offline mobile applications. In this paper we consider several compression techniques for recurrent neural networks including Long–Short Term Memory models. We make particular attention to the high-dimensional output problem caused by the very large vocabulary size. We focus on effective compression methods in the context of their exploitation on devices: pruning, quantization, and matrix decomposition approaches (low-rank factorization and tensor train decomposition, in particular). For each model we investigate the trade-off between its size, suitability for fast inference and perplexity. We propose a general pipeline for applying the most suitable methods to compress recurrent neural networks for language modeling. It has been shown in the experimental study with the Penn Treebank (PTB) dataset that the most efficient results in terms of speed and compression–perplexity balance are obtained by matrix decomposition techniques. [Display omitted] •We compress LSTM networks for word level language modeling up to 5 times.•Low-rank factorization provides the best performance of recurrent neural networks.•Inference in the compressed models can be implemented in real-time on the GPU for a mobile device.•A pipeline to compress recurrent neural networks for language modeling.