Authors
Suriya Gunasekar, Jason D Lee, Daniel Soudry, Nati Srebro
Publication date
2018
Journal
Advances in neural information processing systems
Volume
31
Description
We show that gradient descent on full-width linear convolutional networks of depth converges to a linear predictor related to the 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.
Scholar articles
S Gunasekar, JD Lee, D Soudry, N Srebro - Advances in neural information processing systems, 2018