← Search

Tony Butler-Yeoman

2 accepted papers

2017

Neural Taylor Approximations: Convergence and Exploration in Rectifier Networks

ICLR 2017workshop

Modern convolutional networks, incorporating rectifiers and max-pooling, are neither smooth nor convex. Standard guarantees therefore do not apply. Nevertheless, methods from convex optimization such as gradient descent and Adam are widely used as building blocks for deep learning algorithms. This p…

Cited by 41SourceScholar
2017

Neural Taylor Approximations: Convergence and Exploration in Rectifier Networks

ICML 2017poster

Modern convolutional networks, incorporating rectifiers and max-pooling, are neither smooth nor convex; standard guarantees therefore do not apply. Nevertheless, methods from convex optimization such as gradient descent and Adam are widely used as building blocks for deep learning algorithms. This p…

Cited by 41SourcePDFScholar