AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Large Memory Layers with Product Keys

Guillaume Lample, Alexandre Sablayrolles, Marc'Aurelio Ranzato, Ludovic Denoyer, Herve Jegou

NeurIPS 2019 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 66 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
Add product-key memory to a 12-layer, 1024-dimension TransformerTransformer · CC-News testperplexity: 17.6566 -> 15.6184Improved
Add product-key memory to a 12-layer, 1600-dimension TransformerTransformer · CC-News testperplexity: 14.9994 -> 13.6522Improved
Add query BatchNorm to a memory of size 16384 (ablation)Product-key memory · 6-layer transformer memory ablationperplexity: 22.7972 -> 22.9658Hurt
Add query BatchNorm to a memory of size 262144 (ablation)Product-key memory · 6-layer transformer memory ablationperplexity: 20.4776 -> 19.7805Improved
Add query BatchNorm to a memory of size 1048576 (ablation)Product-key memory · 6-layer transformer memory ablationperplexity: 19.7913 -> 17.9819Improved
Increase memory size from 16384 to 1048576 (ablation)Product-key memory · 6-layer transformer memory ablationperplexity: 22.7972 -> 17.9819Improved
Place memory at intermediate layer 5 rather than layer 1 (ablation)6-layer transformer · 6-layer transformer memory ablationperplexity: 21.4669 -> 19.7805Improved
Replace flat keys with product keys (ablation)Memory layer · Memory size 16384perplexity: 23.2152 -> 22.9658Improved
Replace flat keys with product keys (ablation)Memory layer · Memory size 65536perplexity: 22.5859 -> 21.9114Improved
Replace flat keys with product keys (ablation)Memory layer · Memory size 147456perplexity: 22.1477 -> 20.7050Improved
Replace flat keys with product keys (ablation)Memory layer · Memory size 147456inference speed (w/s): 13.9k -> 36.4kImproved
Increase the number of memory heads or selected neighbors (ablation)Product-key memory · 6-layer transformer memory ablationperplexityImproved

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