AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

GroupReduce: Block-Wise Low-Rank Approximation for Neural Language Model Shrinking

Patrick Chen, Si Si, Yang Li, Ciprian Chelba, Cho-Jui Hsieh

NeurIPS 2018 · Architectures and efficiency › RNNs and normalization · approach: Non-autoregressive translation · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace standard low-rank approximation with GroupReduceNeural language model embeddings · OBWPPL(before retrain): 39.41 -> 32.47Improved
Replace pruning with GroupReduceNeural language model embeddings · OBWPPL(before retrain): 128.31 -> 32.47Improved
Weight approximation by token frequency (ablation)Vanilla SVD for PTB-Small embeddings · PTB-SmallPPL: 189.7 -> 179.8Improved
Use separate low-rank approximations for blocks (ablation)Weighted low-rank PTB-Small embeddings · PTB-SmallPPL: 179.8 -> 155.3Improved
Assign more rank to higher-frequency blocks (ablation)Block low-rank PTB-Small embeddings · PTB-SmallPPL: 155.3 -> 129.2Improved
Refine block assignments using reconstruction error (ablation)Dynamic-rank block approximation for PTB-Small · PTB-SmallPPL: 129.2 -> 127.5Improved
Apply quantization to GroupReduce factorsGroupReduce embeddings · PTB-SmallEmbedding Memory compression: 5x -> 40xImproved
Apply quantization to GroupReduce factorsGroupReduce embeddings · OBWEmbedding Memory compression: 6.6x -> 26xImproved
Vary the number of clusters from 5 to 10, 20, or 30 (ablation)GroupReduce · PTB-LargePPL(before retrain): 81.79; 80.52; 82.88; 83.1Mixed
Use GroupReduce rather than deep compositional codingPTB-Large input embedding · PTB-LargeEmbedding Memory compression: 11.8x -> 23.6xImproved
Use GroupReduce rather than deep compositional codingNMT input embedding · IWSLT 2014 German-to-EnglishEmbedding Memory compression: 16.6x -> 33.3xImproved
Apply low-rank approximation to OBW LSTM cellsOBW-bigLSTM · OBWPPLHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy