Dataset Condensation Atlas

Analysis & theory

Hypergradient

Gradient-based Hyperparameter Optimization through Reversible Learning

Dougal Maclaurin, David Duvenaud, Ryan P. Adams

ICML 2015 · first public 2015-02-11 · arXiv 1502.03492

paper ↗code ↗catalogued✓ abstract read

In one paragraph

Computes exact gradients of validation performance with respect to thousands of hyperparameters — including a data-augmentation network whose weights are treated as hyperparameters — by exactly reversing SGD-with-momentum dynamics, a hypergradient technique later reused to differentiate through the training procedure in bi-level dataset distillation.

Where it sits

Abstract (verbatim from arXiv)

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.

BibTeX (generated; prefer the venue's official entry)
@article{maclaurin2015gradient,
  title   = {Gradient-based Hyperparameter Optimization through Reversible Learning},
  author  = {Dougal Maclaurin and David Duvenaud and Ryan P. Adams},
  journal = {ICML 2015},
  year    = {2015}
}