Neural Networks are Convex Regularizers: Exact Polynomial-time Convex Optimization Formulations for Two-layer Networks

Mert Pilanci · Tolga Ergen

Keywords: [ Computational Learning Theory ] [ Convex Optimization ] [ Non-convex Optimization ] [ Sparsity and Compressed Sensing ] [ Optimization - Convex ]

[ Abstract ]
Wed 15 Jul 10 a.m. PDT — 10:45 a.m. PDT
Wed 15 Jul 11 p.m. PDT — 11:45 p.m. PDT

Abstract: We develop exact representations of two-layer neural networks with rectified linear units in terms of a single convex program with number of variables polynomial in the number of training samples and number of hidden neurons. Our theory utilizes semi-infinite duality and minimum norm regularization. Moreover, we show that certain standard convolutional linear networks are equivalent to $\ell_1$ regularized linear models in a polynomial sized discrete Fourier feature space.

Chat is not available.