LGMLNov 21, 2022

Linear Stability Hypothesis and Rank Stratification for Nonlinear Models

arXiv:2211.11623v19 citationsh-index: 17
Originality Highly original
AI Analysis

This provides a unified framework for understanding target recovery in overparameterized models, which is a foundational problem in machine learning theory.

The paper tackles the mystery of good generalization in overparameterized nonlinear models like deep neural networks by proposing a rank stratification to define an 'effective size of parameters' and establishing a linear stability theory that predicts minimal training data size for target recovery. Experiments show that model rank for specific functions can be much lower than parameter size, enabling recovery even at heavy overparameterization.

Models with nonlinear architectures/parameterizations such as deep neural networks (DNNs) are well known for their mysteriously good generalization performance at overparameterization. In this work, we tackle this mystery from a novel perspective focusing on the transition of the target recovery/fitting accuracy as a function of the training data size. We propose a rank stratification for general nonlinear models to uncover a model rank as an "effective size of parameters" for each function in the function space of the corresponding model. Moreover, we establish a linear stability theory proving that a target function almost surely becomes linearly stable when the training data size equals its model rank. Supported by our experiments, we propose a linear stability hypothesis that linearly stable functions are preferred by nonlinear training. By these results, model rank of a target function predicts a minimal training data size for its successful recovery. Specifically for the matrix factorization model and DNNs of fully-connected or convolutional architectures, our rank stratification shows that the model rank for specific target functions can be much lower than the size of model parameters. This result predicts the target recovery capability even at heavy overparameterization for these nonlinear models as demonstrated quantitatively by our experiments. Overall, our work provides a unified framework with quantitative prediction power to understand the mysterious target recovery behavior at overparameterization for general nonlinear models.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes