In plain words: By exactly rewinding the training run, this technique computes how each setting—like learning rate, initialization, regularization, or architecture—changes the final validation score. That gives exact gradients for thousands of settings at once, where usual tuning guesses one at a time.
Abstract · Gradient-based Hyperparameter Optimization through Reversible Learning
Tuning hyperparameters of learning algorithms is hard because gradients are usually unavailable. We compute exact gradients of cross-validation performance with respect to all hyperparameters by chaining derivatives backwards through the entire training procedure. These gradients allow us to optimize thousands of hyperparameters, including step-size and momentum schedules, weight initialization distributions, richly parameterized regularization schemes, and neural network architectures. We compute hyperparameter gradients by exactly reversing the dynamics of stochastic gradient descent with momentum.
Dougal Maclaurin, David Duvenaud, Ryan P. Adams
arXiv:1502.03492 · stat.ML, cs.LG · submitted Feb 11, 2015 · updated Apr 2, 2015
abstract · pdf · html · 10 figures. Submitted to ICML
Comedy gold.