Efficient and Effective Augmentation Strategy for Adversarial Training

Authors: Sravanti Addepalli, Samyak Jain, Venkatesh Babu R

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We obtain improved robustness and large gains in standard accuracy on multiple datasets (CIFAR-10, CIFAR-100, Image Nette) and model architectures (RN-18, WRN-34-10).
Researcher Affiliation Academia Sravanti Addepalli Samyak Jain R.Venkatesh Babu Video Analytics Lab, Indian Institute of Science, Bangalore Indian Institute of Technology (BHU) Varanasi
Pseudocode Yes (Ref: Algorithm-1 in the Appendix)
Open Source Code Yes The code for implementing DAJAT is available here: https://github.com/val-iisc/DAJAT.
Open Datasets Yes We obtain improved robustness and large gains in standard accuracy on multiple datasets (CIFAR-10, CIFAR-100, Image Nette) and model architectures (RN-18, WRN-34-10).
Dataset Splits Yes Did you specify all the training details (e.g., data splits, hyperparameters, how they were chosen)? [Yes] Appendix-F.2 and Read Me file at https://github.com/val-iisc/DAJAT
Hardware Specification Yes Training time per epoch is reported by running each algorithm across 2 V100 GPUs.
Software Dependencies No The paper mentions software like 'PyTorch' in its references (e.g., [31]) but does not provide specific version numbers for any ancillary software or libraries used in its experiments within the main text or supplementary sections accessible without external links.
Experiment Setup Yes We train all models for 110 epochs unless specified otherwise.