NeurIPS 2018poster493 citations

Implicit Bias of Gradient Descent on Linear Convolutional Networks

Suriya Gunasekar, Jason Lee, Daniel Soudry, Nati Srebro

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.

BibTeX
@inproceedings{NEURIPS2018_0e98aeeb,
 author = {Gunasekar, Suriya and Lee, Jason D and Soudry, Daniel and Srebro, Nati},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {S. Bengio and H. Wallach and H. Larochelle and K. Grauman and N. Cesa-Bianchi and R. Garnett},
 pages = {},
 publisher = {Curran Associates, Inc.},
 title = {Implicit Bias of Gradient Descent on Linear Convolutional Networks},
 url = {https://proceedings.neurips.cc/paper_files/paper/2018/file/0e98aeeb54acf612b9eb4e48a269814c-Paper.pdf},
 volume = {31},
 year = {2018}
}