AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

CATE: Computation-aware Neural Architecture Encoding with Transformers

Shen Yan, Kaiqiang Song, Fei Liu, Mi Zhang

ICML 2021 · Architectures and efficiency › Neural architecture search · approach: Latency-aware architecture search · 10 citations (OpenAlex)

Read the paper

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
Combine CATE-DNGO with local searchCATE-DNGO · NAS-Bench-101 and NAS-Bench-301test error (%): Prev. SOTA: 5.92 -> 5.88; 5.35 -> 5.28Improved
Use a larger query budgetCATE-DNGO-LS · DARTS search space (CIFAR-10)average test error (%): 2.55 ± 0.08 -> 2.46 ± 0.05Improved
Use an indirect dependency mask to capture long-range dependencies (ablation)CATE-DNGO · NAS-Bench-101 and NAS-Bench-301test error (%): 6.03 -> 5.94; 5.35 -> 5.30Improved
Use unidirectional rather than bidirectional encoding (ablation)CATE-DNGO-LS · NAS-Bench-101 and NAS-Bench-301test error (%): 5.89 -> 5.88; 5.30 -> 5.28Improved
Use a smaller computational-neighborhood threshold δ (ablation)CATE-DNGO · NAS-Bench-101test error (%): δ=2×10^6, K=2: 5.94; δ=8×10^6, K=4: 6.11Mixed
Vary the neighborhood size K (ablation)CATE-DNGO · NAS-Bench-101test error (%): 5.94 to 6.11No clear effect
Increase the number of cross-attention Transformer blocks and feed-forward dimension (ablation)CATE-DNGO · NAS-Bench-101test error (%): 6.07 -> 5.94Improved
Increase the corruption rate from 20% to 30% (ablation)CATE · NAS-Bench-101 and NAS-Bench-301test error (%): 5.88 -> 5.93; 5.28 -> 5.29Hurt
Increase the corruption rate from 15% to 20% (ablation)CATE · NAS-Bench-101test error (%): 5.89 -> 5.88Improved
Use a computation-aware encoding in the outside search spaceCATE predictor · NAS-Bench-101 outside search spacevalidation errorImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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