Research map / Architectures and efficiency / Neural architecture search
Does Unsupervised Architecture Representation Learning Help Neural Architecture Search?
Shen Yan, Yu Zheng, Wei Ao, Xiao Zeng, Mi Zhang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| learn architecture embeddings unsupervised (decoupled from search) instead of supervised joint learning | arch2vec RL search · NAS-Bench-101 | test accuracy (%): 93.74 -> 94.10 | Improved |
| learn embeddings unsupervised instead of supervised | arch2vec BO search · NAS-Bench-101 | test accuracy (%): 93.79 -> 94.05 | Improved |
| add KL regularization term to the autoencoder (ablation) | arch2vec · NAS-Bench-101 validity | validity (%): 30.31 -> 44.97 | Improved |
| add KL regularization term to the autoencoder (ablation) | arch2vec · DARTS validity | validity (%): 16.01 -> 33.36 | Improved |
| add KL regularization term to the autoencoder (ablation) | arch2vec · NAS-Bench-201 reconstruction accuracy | reconstruction accuracy: 100 -> 100 | No clear effect |
| unsupervised embeddings instead of supervised for GP performance prediction | arch2vec · NAS-Bench-101 | RMSE: 0.038 -> 0.018 | Improved |
| use arch2vec embeddings with BO instead of discrete RE | NAS-Bench-201 · CIFAR-100 | test accuracy (%): 72.86 -> 73.37 | Improved |
| arch2vec-BO vs supervised BANANAS in DARTS space | DARTS search space · CIFAR-10 | avg test error (%): 2.67 -> 2.56 | Improved |
| discrete adjacency-matrix encoding replaced by supervised continuous embeddings | RE/BOHB search · NAS-Bench-101 | test regret | Improved |
Closest papers by meaning
- Are Labels Necessary for Neural Architecture Search? (ECCV 2020 · similarity 0.90)
- Understanding Architectures Learnt by Cell-based Neural Architecture Search (ICLR 2020 · similarity 0.90)
- Neural Architecture Search With Representation Mutual Information (CVPR 2022 · similarity 0.89)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.89)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.89)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · similarity 0.88)
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.
