preprint · arXiv (Cornell University) · 2023

[Re] Double Sampling Randomized Smoothing

Aryan Gupta, Sarthak Gupta, Abhay Kumar, Harsh Dugar

Cited by 0Open access

Abstract

This paper is a contribution to the reproducibility challenge in the field of machine learning, specifically addressing the issue of certifying the robustness of neural networks (NNs) against adversarial perturbations. The proposed Double Sampling Randomized Smoothing (DSRS) framework overcomes the limitations of existing methods by using an additional smoothing distribution to improve the robustness certification. The paper provides a clear manifestation of DSRS for a generalized family of Gaussian smoothing and a computationally efficient method for implementation. The experiments on MNIST and CIFAR-10 demonstrate the effectiveness of DSRS, consistently certifying larger robust radii compared to other methods. Also various ablations studies are conducted to further analyze the hyperparameters and effect of adversarial training methods on the certified radius by the proposed framework.

Cite this paper

APA 7

Gupta, A., Gupta, S., Kumar, A., & Dugar, H. (2023). [Re] Double Sampling Randomized Smoothing. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2306.15221

MLA 9

Gupta, Aryan, et al. “[Re] Double Sampling Randomized Smoothing.” arXiv (Cornell University), 2023, https://doi.org/10.48550/arxiv.2306.15221.

Chicago

Gupta, Aryan, Sarthak Gupta, Abhay Kumar, and Harsh Dugar. 2023. “[Re] Double Sampling Randomized Smoothing.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2306.15221.

Harvard

Gupta, A. et al. (2023) ‘[Re] Double Sampling Randomized Smoothing’, arXiv (Cornell University). Available at: https://doi.org/10.48550/arxiv.2306.15221.

BibTeX

@misc{gupta2023double,
  title = {{[Re] Double Sampling Randomized Smoothing}},
  author = {Gupta, Aryan and Gupta, Sarthak and Kumar, Abhay and Dugar, Harsh},
  journal = {arXiv (Cornell University)},
  year = {2023},
  doi = {10.48550/arxiv.2306.15221}
}

Made from the index record. Check names, capitalisation and page numbers against the paper. Need another style? AI for Research has over 10,000.