Research map / Architectures and efficiency / Neural architecture search
Learnable Embedding Space for Efficient Neural Architecture Compression
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-100, VGG-19 teacher | accuracy: 68.17% -> 71.41% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-100, ResNet-18 teacher | accuracy: 69.86% -> 73.83% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-100, ResNet-34 teacher | accuracy: 72.33% -> 73.68% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-100, ShuffleNet teacher | accuracy: 64.75% -> 68.45% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-10, VGG-19 teacher | accuracy: 91.76% -> 92.27% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-10, ResNet-18 teacher | accuracy: 92.29% -> 92.99% | Helped |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-10, ResNet-34 teacher | accuracy: 92.87% -> 92.70% | Hurt |
| use the learned embedding with Bayesian optimization rather than random search | compressed architecture search · CIFAR-10, ShuffleNet teacher | accuracy: 88.25% -> 89.36% | Helped |
| add multiple kernels rather than using one kernel (ablation) | architecture search with ResNet-34 teacher · CIFAR-100 | f(x): 0.9181; 0.9119; 0.9238; 0.9240; 0.9256 | Mixed |
| add the option to create skip connections (ablation) | architecture search · CIFAR-100, ResNet-18 teacher | f(x): 0.9062 -> 0.9123 | Helped |
| add the option to create skip connections (ablation) | architecture search · CIFAR-100, ResNet-34 teacher | f(x): 0.9205 -> 0.9246 | Helped |
| use evolutionary search rather than random sampling to maximize the acquisition function (ablation) | architecture search with one kernel · CIFAR-100, ResNet-34 teacher | f(x): 0.9181 -> 0.9056 | Hurt |
Closest papers by meaning
- ISTA-NAS: Efficient and Consistent Neural Architecture Search by Sparse Coding (NeurIPS 2020 · similarity 0.88)
- Interpretable Neural Architecture Search via Bayesian Optimisation with Weisfeiler-Lehman Kernels (ICLR 2021 · similarity 0.87)
- BayesNAS: A Bayesian Approach for Neural Architecture Search (ICML 2019 · similarity 0.87)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.85)
- einspace: Searching for Neural Architectures from Fundamental Operations (NeurIPS 2024 · similarity 0.85)
- Efficient Few-Shot Neural Architecture Search by Counting the Number of Nonlinear Functions (AAAI 2025 · similarity 0.84)
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.