Research map / Optimization / Implicit bias and generalization
Greedy Pruning with Group Lasso Provably Generalizes for Matrix Sensing
Nived Rajaraman, Fnu Devvrit, Aryan Mokhtari, Kannan Ramchandran
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 smooth group Lasso regularizer to training loss (ablation) | gradient descent on overparameterized matrix sensing · matrix sensing, d=k=500, r=4 | column-norm sparsity (fraction of large columns) | Improved |
| train without regularization (implicit regularization from small initialization) (ablation) | gradient descent · matrix sensing, d=k=500, r=4 | column-norm sparsity (suitability for pruning) | Hurt |
| greedy pruning of small-norm columns then fine-tuning, versus vanilla gradient descent (ablation) | group Lasso training · matrix sensing, d=k=20, r=3, n=100 noiseless | generalization error ||U U^T - U* U*^T||_F^2 | Improved |
| prune columns below the l2-norm threshold after regularized training (ablation) | group Lasso trained model · matrix sensing, d=k=20, r=3, n=100 | generalization error | Improved |
Closest papers by meaning
- SequentialAttention++ for Block Sparsification: Differentiable Pruning Meets Combinatorial Optimization (NeurIPS 2024 · similarity 0.89)
- Sparsest Models Elude Pruning: An Exposé of Pruning’s Current Capabilities (ICML 2024 · similarity 0.88)
- Targeted Low-rank Refinement: Enhancing Sparse Language Models with Precision (ICML 2025 · similarity 0.87)
- Two Sparse Matrices are Better than One: Sparsifying Neural Networks with Double Sparse Factorization (ICLR 2025 · similarity 0.87)
- How Sparse Can We Prune A Deep Network: A Fundamental Limit Perspective (NeurIPS 2024 · similarity 0.86)
- Compress to Impress: Efficient LLM Adaptation Using a Single Gradient Step on 100 Samples (NeurIPS 2025 · similarity 0.86)
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.
