Poster

Implicit Bias of Gradient Descent on Linear Convolutional Networks

Suriya Gunasekar ⋅ Jason Lee ⋅ Daniel Soudry ⋅ Nati Srebro

Keywords: Non-Convex Optimization Optimization for Deep Networks CNN Architectures

2018 Poster

[ Paper]

Abstract

We show that gradient descent on full-width linear convolutional networks of depth $L$ converges to a linear predictor related to the $\ell_{2/L}$ bridge penalty in the frequency domain. This is in contrast to linearly fully connected networks, where gradient descent converges to the hard margin linear SVM solution, regardless of depth.

Chat is not available.