Poster

Implicit Convex Regularizers of CNN Architectures: Convex Optimization of Two- and Three-Layer Networks in Polynomial Time

Tolga Ergen · Mert Pilanci

Virtual

Keywords: [ polynomial time ] [ convex duality ] [ $\ell_1$ norm ] [ group sparsity ] [ convex optimization ] [ non-convex optimization ] [ deep learning ]

[ Abstract ]
[ Slides
[ Paper ]
Tue 4 May 5 p.m. PDT — 7 p.m. PDT
 
Spotlight presentation: Oral Session 5
Tue 4 May 11 a.m. PDT — 1:56 p.m. PDT

Abstract: We study training of Convolutional Neural Networks (CNNs) with ReLU activations and introduce exact convex optimization formulations with a polynomial complexity with respect to the number of data samples, the number of neurons, and data dimension. More specifically, we develop a convex analytic framework utilizing semi-infinite duality to obtain equivalent convex optimization problems for several two- and three-layer CNN architectures. We first prove that two-layer CNNs can be globally optimized via an $\ell_2$ norm regularized convex program. We then show that multi-layer circular CNN training problems with a single ReLU layer are equivalent to an $\ell_1$ regularized convex program that encourages sparsity in the spectral domain. We also extend these results to three-layer CNNs with two ReLU layers. Furthermore, we present extensions of our approach to different pooling methods, which elucidates the implicit architectural bias as convex regularizers.

Chat is not available.