Trusting SVM for Piecewise Linear CNNs

Leonard Berrada, Andrew Zisserman, M. Pawan Kumar
11/7/2016
cs.LG

Official code: oval-group/pl-cnn (8 stars, MIT, last commit 2017-05-09).

Abstract

We present a novel layerwise optimization algorithm for the learning objective of Piecewise-Linear Convolutional Neural Networks (PL-CNNs), a large class of convolutional neural networks. Specifically, PL-CNNs employ piecewise linear non-linearities such as the commonly used ReLU and max-pool, and an SVM classifier as the final layer. The key observation of our approach is that the problem corresponding to the parameter estimation of a layer can be formulated as a difference-of-convex (DC) program, which happens to be a latent structured SVM. We optimize the DC program using the concave-convex procedure, which requires us to iteratively solve a structured SVM problem. This allows to design an optimization algorithm with an optimal learning rate that does not require any tuning. Using the MNIST, CIFAR and ImageNet data sets, we show that our approach always improves over the state of the art variants of backpropagation and scales to large data and large network settings.

View on arXivView PDF

Code Implementations(1)

oval-group/pl-cnnOfficial100%
MIT

A layer-wise algorithm for learning Convolutional Neural Networks

86Shell, PythonFeb 18, 20179 years agoMIT

Similar Papers

A Globally Convergent Algorithm for Neural Network Parameter Optimization Based on Difference-of-Convex Functions

84% match2024

DFWLayer: Differentiable Frank-Wolfe Optimization Layer

83% match1 repo22023

Morphological Perceptron with Competitive Layer: Training Using Convex-Concave Procedure

83% match2025

The Convex Landscape of Neural Networks: Characterizing Global Optima and Stationary Points via Lasso Models

82% match1 repo192023

Convolutional Channel-wise Competitive Learning for the Forward-Forward Algorithm

82% match1 repo132023

Karush-Kuhn-Tucker Condition-Trained Neural Networks (KKT Nets)

82% match2024

Cite this paper

@article{berrada2016trusting,
  title  = {Trusting SVM for Piecewise Linear CNNs},
  author = {Leonard Berrada and Andrew Zisserman and M. Pawan Kumar},
  year   = {2016},
  eprint = {1611.02185},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1611.02185}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1611.02185.svg)](https://codeofpaper.com/paper/1611.02185)

Discussion