Scaling provable adversarial defenses

Authors: Eric Wong, Frank Schmidt, Jan Hendrik Metzen, J. Zico Kolter

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On both MNIST and CIFAR data sets, we train classifiers that improve substantially on the state of the art in provable robust adversarial error bounds: from 5.8% to 3.1% on MNIST (with ℓ perturbations of ϵ = 0.1), and from 80% to 36.4% on CIFAR (with ℓ perturbations of ϵ = 2/255).
Researcher Affiliation Collaboration Eric Wong Machine Learning Department Carnegie Mellon University Pittsburgh, PA 15213 ericwong@cs.cmu.edu Frank R. Schmidt Bosch Center for Artificial Intelligence Renningen, Germany frank.r.schmidt@de.bosch.com Jan Hendrik Metzen Bosch Center for Artificial Intelligence Renningen, Germany janhendrik.metzen@de.bosch.com J. Zico Kolter Computer Science Department Carnegie Mellon University and Bosch Center for Artificial Intelligence Pittsburgh, PA 15213 zkolter@cs.cmu.edu
Pseudocode Yes Algorithm 1 Estimating ν1 1 and P j I ℓij[νij]+
Open Source Code Yes Code for all experiments in the paper is available at https://github.com/locuslab/convex_adversarial/.
Open Datasets Yes We evaluate the techniques in this paper on two main datasets: MNIST digit classification [Le Cun et al., 1998] and CIFAR10 image classification [Krizhevsky, 2009].
Dataset Splits No The paper uses MNIST and CIFAR10 datasets but does not explicitly provide percentages or counts for train/validation/test splits, nor does it reference a specific, predefined split that includes a validation set. It discusses training and testing errors, but a separate validation set split is not detailed.
Hardware Specification Yes Each training epoch with 10 random projections takes less than a minute on a single Ge Force GTX 1080 Ti graphics card, while using less than 700MB of memory
Software Dependencies No The paper does not provide specific version numbers for any software dependencies or libraries used in the experiments.
Experiment Setup Yes The ϵ value for training is scheduled from 0.01 to 0.1 over the first 20 epochs.