PAC-Bayesian Margin Bounds for Convolutional Neural Networks
arXiv:1801.00171v226 citations
Originality Synthesis-oriented
AI Analysis
This work addresses the generalization error problem for researchers in machine learning theory, but it is incremental as it extends an existing framework to a specific architecture.
The authors tackled the problem of analyzing generalization error in convolutional neural networks by adapting the PAC-Bayesian framework from fully connected layers, resulting in theoretical bounds for this setting.
Recently the generalization error of deep neural networks has been analyzed through the PAC-Bayesian framework, for the case of fully connected layers. We adapt this approach to the convolutional setting.