Improved, Deterministic Smoothing for L_1 Certified Robustness

Authors: Alexander J Levine, Soheil Feizi

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental On CIFAR-10 and Image Net datasets, we provide substantially larger ℓ1 robustness certificates compared to prior works, establishing a new state-ofthe-art. The determinism of our method also leads to significantly faster certificate computation. Code is available at: https://github.com/ alevine0/smoothing Splitting Noise. ... We evaluated the performance of our method on CIFAR-10 and Image Net datasets, matching all experimental conditions from (Yang et al., 2020) as closely as possible (further details are given in the appendix.) Certification performance data is given in Table 1 for CIFAR-10 and Figure 7 for Imagenet.
Researcher Affiliation Academia Alexander Levine 1 Soheil Feizi 1 ... 1Department of Computer Science, University of Maryland, College Park, Maryland, USA.
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at: https://github.com/ alevine0/smoothing Splitting Noise.
Open Datasets Yes On CIFAR-10 and Image Net datasets, we provide substantially larger ℓ1 robustness certificates compared to prior works
Dataset Splits No The paper mentions using CIFAR-10 and ImageNet datasets and matching experimental conditions from previous work, but it does not explicitly provide specific training, validation, or test dataset split percentages or sample counts within the main text.
Hardware Specification Yes We used a single NVIDIA 2080 Ti GPU.
Software Dependencies No The paper does not provide specific software details, such as library or solver names with version numbers.
Experiment Setup No The paper mentions matching experimental conditions from a previous work (Yang et al., 2020) and that further details are in the appendix, but it does not provide concrete hyperparameter values or detailed training configurations within the main text.