We propose a stepsize adaptation scheme for stochastic gradient descent. It operates directly with the loss function and rescales the gradient in order to make fixed predicted progress on the loss. We demonstrate its capabilities by conclusively improving the performance of Adam and Momentum optimizers. The enhanced optimizers with default hyperparameters consistently outperform their constant stepsize counterparts, even the best ones, without a measurable increase in computational cost. The performance is validated on multiple architectures including dense nets, CNNs, ResNets, and the recurrent Differential Neural Computer on classical datasets MNIST, fashion MNIST, CIFAR10 and others.
Michal Rolinek (Max Planck Institute for Intelligent Systems)
Georg Martius (MPI for Intelligent Systems)
More from the Same Authors
2020 Workshop: Learning Meets Combinatorial Algorithms »
Marin Vlastelica · Jialin Song · Aaron Ferber · Brandon Amos · Georg Martius · Bistra Dilkina · Yisong Yue
2019 Poster: Control What You Can: Intrinsically Motivated Task-Planning Agent »
Sebastian Blaes · Marin Vlastelica Pogančić · Jiajie Zhu · Georg Martius