Research map / Label-efficient and robust learning / Recognition and long-tail
Where to Pay Attention in Sparse Training for Feature Selection?
Ghada Sokar, Zahra Atashgahi, Mykola Pechenizkiy, Decebal Constantin Mocanu
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 |
|---|---|---|---|
| add loss-gradient sensitivity to neuron importance (ablation) | WAST · Madelon | accuracy (%): 55.50 -> 83.27 | Improved |
| add loss-gradient sensitivity to neuron importance (ablation) | WAST · PCMAC | accuracy (%): 55.58 -> 60.51 | Improved |
| add loss-gradient sensitivity to neuron importance (ablation) | WAST · HAR | accuracy (%): 87.88 -> 91.20 | Improved |
| add loss-gradient sensitivity to neuron importance (ablation) | WAST · SMK | accuracy (%): 84.21 -> 84.74 | No clear effect |
| add weight-magnitude term to neuron importance (ablation) | WAST · SMK | accuracy (%): 75.26 -> 84.74 | Improved |
| add weight-magnitude term to neuron importance (ablation) | WAST · HAR | accuracy (%): 92.20 -> 91.20 | Hurt |
| add weight-magnitude term to neuron importance (ablation) | WAST · PCMAC | accuracy (%): 58.66 -> 60.51 | Mixed |
| add momentum (accumulated importance across iterations) (ablation) | WAST · SMK | accuracy (%): 78.95 -> 84.74 | Improved |
| add momentum (accumulated importance across iterations) (ablation) | WAST · USPS | accuracy (%): 96.60 -> 96.69 | No clear effect |
| include neuron importance in connection importance (ablation) | WAST · PCMAC | accuracy (%): 55.53 -> 60.51 | Improved |
| include neuron importance in connection importance (ablation) | WAST · Madelon | accuracy (%): 83.27 -> 83.27 | No clear effect |
| replace random regrowth (QS) with attention-guided regrowth, 10 vs 100 epochs | WAST · Madelon | accuracy (%): 82.07 -> 83.27 | Improved |
Closest papers by meaning
- Sequential Attention for Feature Selection (ICLR 2023 · similarity 0.89)
- SAND: One-Shot Feature Selection with Additive Noise Distortion (ICML 2025 · similarity 0.87)
- Selective Deep Autoencoder for Unsupervised Feature Selection (AAAI 2024 · similarity 0.85)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
- Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency (NeurIPS 2025 · similarity 0.84)
- Estimating Conditional Mutual Information for Dynamic Feature Selection (ICLR 2024 · 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.
