Curse of Dimensionality on Randomized Smoothing for Certifiable Robustness

Authors: Aounon Kumar, Alexander Levine, Tom Goldstein, Soheil Feizi

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present experimental results on CIFAR to validate our theory. For other smoothing distributions, such as, a uniform distribution within an ℓ1 or an ℓ∞-norm ball, we show upper bounds of the form O(1/d) and O(1/d1−1/p) respectively, which have an even worse dependence on d.
Researcher Affiliation Academia 1University of Maryland, College Park, Maryland, USA. Correspondence to: Aounon Kumar <aounon@umd.edu>, Soheil Feizi <sfeizi@cs.umd.edu>.
Pseudocode No The paper does not contain any pseudocode or algorithm blocks.
Open Source Code Yes The code for our experiments is available on Git Hub at: https://github.com/alevine0/ smoothing Gen Gaussian
Open Datasets Yes We present experimental results on CIFAR to validate our theory. We provide empirical evidence to support our claims on the CIFAR-10 dataset.
Dataset Splits No The paper mentions using 100,000 samples for estimating p1(x) and p2(x), but does not explicitly state train/validation/test dataset splits (e.g., percentages or counts) for reproducibility.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments (e.g., GPU models, CPU types, or cloud instance specifications).
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., Python version, library versions like PyTorch, TensorFlow, etc.).
Experiment Setup Yes We specifically tested on CIFAR-10 (32x32 pixels), as well as scaled-down versions of this dataset (16x16 and 8x8 pixels)... Note that we re-trained the classifier on noisy images for each noise distribution and standard deviation σ. For a fixed standard deviation σ... (Figure 5) ... (σ = .12) ... (σ = .25).