Adversarial training objective

A robust objective that evaluates loss on perturbations selected to increase error within a specified threat set.

A robust objective that evaluates loss on perturbations selected to increase error within a specified threat set.

is aObjective All entries Open in graph

The perturbation set, inner maximization, base loss, and aggregation rule determine the robustness notion being optimized.

Connections

optimized by

Adversarial training

specialization of

Loss function