/pytoch

Official source code for "Is Adversarial Training with Compressed Datasets Effective?" (Chen & Selvan. 2024)

Primary LanguagePythonMIT LicenseMIT

Minimal Finite Covering

This repository implements the minimal finite covering problem and its application, and is the official source code for paper "Is Adversarial Training with Compressed Datasets Effective?" (Chen & Selvan. 2024). For more information, contact: Tong Chen (toch@di.ku.dk).

Requirements

torch, numpy, scipy, gurobipy (license is required)

Basic Usage

Finite covering with fixed number $k$ and minimum radius $\eta$.

sub_data, eta, sol = FC(data).gen_data(A, k=k, save=False)

Finite covering with fixed radius $\eta$ and minimum number $k$.

sub_data, eta, sol = FC(data).gen_data(A, eta=eta, save=False)