Implicit Bias of Gradient Descent for Wide Two-layer Neural Networks Trained with the Logistic Loss

Abstract

Neural networks trained to minimize the logistic (a.k.a. cross-entropy) loss with gradient-based methods are observed to perform well in many supervised classification tasks. Towards understanding this phenomenon, we analyze the training and generalization behavior of infinitely wide two-layer neural networks with homogeneous activations. We show that the limits of the gradient flow on exponentially tailed losses can be fully characterized as a max-margin classifier in a certain non-Hilbertian space of functions. In presence of hidden low-dimensional structures, the resulting margin is independent of the ambiant dimension, which leads to strong generalization bounds. In contrast, training only the output layer implicitly solves a kernel support vector machine, which a priori does not enjoy such an adaptivity. Our analysis of training is non-quantitative in terms of running time but we prove computational guarantees in simplified settings by showing equivalences with online mirror descent. Finally, numerical experiments suggest that our analysis describes well the practical behavior of two-layer neural networks with ReLU activation and confirm the statistical benefits of this implicit bias.

Cited by in corpus (10)

• The Modern Mathematics of Deep Learning
• Deep Learning Meets Sparse Regularization: A Signal Processing Perspective
• Mathematical Models of Overparameterized Neural Networks
• Understanding Deep Learning via Decision Boundary
• Relative stability toward diffeomorphisms indicates performance in deep nets
• The impact of memory on learning sequence-to-sequence tasks
• Regression as Classification: Influence of Task Formulation on Neural Network Features
• Abide by the Law and Follow the Flow: Conservation Laws for Gradient Flows
• Naive imputation implicitly regularizes high-dimensional linear models
• Is interpolation benign for random forest regression?