资源分类
》 资源 》论文》Implicit Bias of Gradient Descent on Linear Convolutional Networks

Implicit Bias of Gradient Descent on Linear Convolutional Networks

2020-02-17 | |  45 |   35 |   0

Abstract 

We show that gradient descent on full width linear convolutional networks of depth L converges to a linear predictor related to the image.png bridge penalty in the frequency domain. This is in contrast to fully connected linear networks, where regardless of depth, gradient descent converges to the image.png maximum margin solution.

上一篇：Transfer Learning with Neural AutoML

下一篇：GIANT: Globally Improved Approximate Newton Method for Distributed Optimization

用户评价
全部评价

热门资源

  • Learning to Predi...

    Much of model-based reinforcement learning invo...

  • Stratified Strate...

    In this paper we introduce Stratified Strategy ...

  • The Variational S...

    Unlike traditional images which do not offer in...

  • A Mathematical Mo...

    Direct democracy, where each voter casts one vo...

  • Rating-Boosted La...

    The performance of a recommendation system reli...