Research map / Optimization / Implicit bias and generalization
Understanding the Dynamics of Gradient Flow in Overparameterized Linear models
Salma Tarmoun, Guilherme Franca, Benjamin D Haeffele, Rene Vidal
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 |
|---|---|---|---|
| Initialize overparameterized factorization with larger conserved imbalance ||Q0|| (balanced 0.02 -> imbalanced 196.55) | two-layer linear model (asymmetric factorization, gradient descent) · random Gaussian Y, m=5, n=10, k=50, synthetic | reconstruction error convergence speed | Improved |
| Increase imbalance of initialization (||Q0|| from 0.02 to 1.83) (ablation) | two-layer linear model · random Gaussian Y, m=5, n=10, k=50, synthetic | convergence speed of singular values | Improved |
| Use balanced (near-zero conserved quantity) initialization instead of imbalanced (ablation) | two-layer linear model · random Gaussian Y, synthetic | order of singular value learning | No clear effect |
| Add imbalance-driven conservation law (nonzero Q0) to nonlinear two-layer sigmoid network | nonlinear two-layer network · synthetic data, n=1000, d=10, width 100 | training loss | No clear effect |
| Replace the 1/8 weight of the regularizer in explicit-regularization dynamics by a different constant alpha | regularized factorization dynamics · theory | equivalence to conservation-law dynamics | Hurt |
Closest papers by meaning
- On the Convergence of Gradient Flow on Multi-layer Linear Models (ICML 2023 · similarity 0.91)
- Random Matrix Theory of Early-Stopped Gradient Flow: A Transient BBP Scenario (NeurIPS 2026 · similarity 0.88)
- Keep the Momentum: Conservation Laws beyond Euclidean Gradient Flows (ICML 2024 · similarity 0.87)
- Transformative or Conservative? Conservation laws for ResNets and Transformers (ICML 2025 · similarity 0.86)
- Mixed Dynamics In Linear Networks: Unifying the Lazy and Active Regimes (NeurIPS 2024 · similarity 0.86)
- Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares (NeurIPS 2025 · similarity 0.85)
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.
