Guaranteed Recovery of One-Hidden-Layer Neural Networks via Cross Entropy

TL;DR

This paper proves that gradient descent can reliably recover the weights of a one-hidden-layer neural network from Gaussian inputs using cross entropy, with guarantees on convergence, sample complexity, and initialization.

Contribution

It establishes the first global convergence guarantees for empirical risk minimization with cross entropy in one-hidden-layer neural networks, including initialization and sample complexity analysis.

Findings

Gradient descent converges linearly to the ground truth when initialized properly.

Empirical risk exhibits strong convexity and smoothness near the ground truth.

The tensor method provides a valid initialization for the training process.

Abstract

We study model recovery for data classification, where the training labels are generated from a one-hidden-layer neural network with sigmoid activations, also known as a single-layer feedforward network, and the goal is to recover the weights of the neural network. We consider two network models, the fully-connected network (FCN) and the non-overlapping convolutional neural network (CNN). We prove that with Gaussian inputs, the empirical risk based on cross entropy exhibits strong convexity and smoothness {\em uniformly} in a local neighborhood of the ground truth, as soon as the sample complexity is sufficiently large. This implies that if initialized in this neighborhood, gradient descent converges linearly to a critical point that is provably close to the ground truth. Furthermore, we show such an initialization can be obtained via the tensor method. This establishes the global convergence guarantee for empirical risk minimization using cross entropy via gradient descent for learning one-hidden-layer neural networks, at the near-optimal sample and computational complexity with respect to the network input dimension without unrealistic assumptions such as requiring a fresh set of samples at each iteration.