Improving Word Embedding Factorization for Compression Using Distilled Nonlinear Neural Decomposition

Abstract

Word-embeddings are vital components of Natural Language Processing (NLP) models and have been extensively explored. However, they consume a lot of memory which poses a challenge for edge deployment. Embedding matrices, typically, contain most of the parameters for language models and about a third for machine translation systems. In this paper, we propose Distilled Embedding, an (input/output) embedding compression method based on low-rank matrix decomposition and knowledge distillation. First, we initialize the weights of our decomposed matrices by learning to reconstruct the full pre-trained word-embedding and then fine-tune end-to-end, employing knowledge distillation on the factorized embedding. We conduct extensive experiments with various compression rates on machine translation and language modeling, using different data-sets with a shared word-embedding matrix for both embedding and vocabulary projection matrices. We show that the proposed technique is simple to replicate, with one fixed parameter controlling compression size, has higher BLEU score on translation and lower perplexity on language modeling compared to complex, difficult to tune state-of-the-art methods.

arXiv commentAccepted at Findings of EMNLP 2020

References in corpus (10)

• Deep Learning in Neural Networks: An Overview
• Distilling the Knowledge in a Neural Network
• Sequence to Sequence Learning with Neural Networks
• Language Models are Few-Shot Learners
• Tensorizing Neural Networks
• Regularizing Neural Networks by Penalizing Confident Output Distributions
• Multilingual Neural Machine Translation with Knowledge Distillation
• Tensorized Embedding Layers for Efficient Model Compression
• On the Effectiveness of Low-Rank Matrix Factorization for LSTM Model Compression
• Slim Embedding Layers for Recurrent Neural Language Models