A Noise Sensitivity Exponent Controls Large Statistical-to-Computational Gaps in Single- and Multi-Index Models
stat.ML, cs.LG
Submitted: 2026-03-18
Updated: 2026-11-16
Terminology
Sources
- Repetita Iuvant: Data Repetition Allows SGD to Learn High-Dimensional Multi-Index Functions
- Learning quadratic neural networks in high dimensions: SGD dynamics and scaling laws
- Generalization performance of narrow one-hidden layer networks in the teacher-student setting
- On Learning Gaussian Multi-index Models with Gradient Flow
- Single-Head Attention in High Dimensions: A Theory of Generalization, Weights Spectra, and Scaling Laws
- Survey on Algorithms for multi-index models
- An Optimized Franz-Parisi Criterion and its Equivalence with SQ Lower Bounds
- Low-dimensional Functions are Efficiently Learnable under Randomly Biased Distributions
- The Generative Leap: Sharp Sample Complexity for Efficiently Learning Gaussian Multi-Index Models
- The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents
- Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime
- Optimal scaling laws in learning hierarchical multi-index models
- A Theory of Non-Linear Feature Learning with One Gradient Step in Two-Layer Neural Networks
- Emergence and scaling laws in SGD learning of shallow neural networks
- Computational Complexity of Statistics: New Insights from Low-Degree Polynomials
Related papers
- Uniform-in-Time Weak Propagation-of-Chaos in Shallow Neural Networks
- Diffusion Models Observe Only Gradients: A Geometric Perspective on Score Matching Errors
- Quantifying Error Propagation and Model Collapse in Diffusion Models
- Behavior of prediction performance metrics with rare events
- Statistical Inference and Learning for Shapley Additive Explanations (SHAP)
- Control Variate Design for Score-Function Gradient Estimators in Importance-Weighted Variational Inference