Research map / Architectures and efficiency / Neural architecture search
Learning from Mistakes – a Framework for Neural Architecture Search
Bhanu Garg, Li Zhang, Pradyumna Sridhara, Ramtin Hosseini, Eric Xing, Pengtao Xie
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 |
|---|---|---|---|
| Apply LFM to DARTS-2nd | DARTS-2nd · CIFAR-100 | test error (%): 20.58\pm0.44 -> 17.65\pm0.45 | Improved |
| Apply LFM to PDARTS with a ResNet-18 encoder | PDARTS · CIFAR-100 | test error (%): 17.49 -> 16.44\pm0.11 | Improved |
| Use a ResNet-34 encoder instead of ResNet-18 (ablation) | LFM-PDARTS · CIFAR-100 | test error (%): 0.75% | Improved |
| Apply LFM to DARTS-2nd | DARTS-2nd · CIFAR-10 | test error (%): 2.76\pm0.09 -> 2.70\pm0.06 | Improved |
| Apply LFM to PDARTS | PDARTS · CIFAR-10 | test error (%): 2.50 -> 2.46\pm0.04 | Improved |
| Apply LFM to DARTS-2nd architecture searched on CIFAR-10 | DARTS2nd-CIFAR10 · ImageNet | top-1 error (%): 26.7 -> 25.12 | Improved |
| Apply LFM to PDARTS architecture searched on CIFAR-10 | PDARTS (CIFAR10) · ImageNet | top-5 error (%): 7.4 -> 6.85 | Improved |
| Use an encoder trained only on CIFAR-100 rather than ImageNet-pretrained (ablation) | LFM+DARTS · CIFAR-100 | test error (%): 17.65\pm0.45 -> 17.82\pm0.39 | No clear effect |
| Use an encoder trained only on CIFAR-100 rather than ImageNet-pretrained (ablation) | LFM+PDARTS · CIFAR-100 | test error (%): 16.44\pm0.11 -> 16.51\pm0.13 | No clear effect |
| Add parameter sharing between learner models and use W1 as the encoder (ablation) | LFM+DARTS · CIFAR-100 | test error (%): 17.65\pm0.45 -> 18.77\pm0.31 | Hurt |
| Add parameter sharing between learner models and use W1 as the encoder (ablation) | LFM+PDARTS · CIFAR-100 | test error (%): 16.44\pm0.11 -> 16.83\pm0.08 | Hurt |
| Use dot-product attention rather than cosine similarity or L2 distance for visual similarity (ablation) | LFM · CIFAR-100 | test error | Improved |
Closest papers by meaning
- Unveiling AI's Blind Spots: An Oracle for In-Domain, Out-of-Domain, and Adversarial Errors (ICML 2025 · similarity 0.85)
- Building Optimal Neural Architectures using Interpretable Knowledge (CVPR 2024 · similarity 0.83)
- Challenging Forgets: Unveiling the Worst-Case Forget Sets in Machine Unlearning (ECCV 2024 · similarity 0.82)
- Insights from the Use of Previously Unseen Neural Architecture Search Datasets (CVPR 2024 · similarity 0.82)
- Revisiting Neural Networks for Few-Shot Learning: A Zero-Cost NAS Perspective (ICML 2025 · similarity 0.82)
- Transformer Copilot: Learning from The Mistake Log in LLM Fine-tuning (NeurIPS 2025 · similarity 0.82)
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.
