SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.02309 · ICML · 2024

Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing

Linyi Li, Bo Li, Xi Xiao, Minhui Xue, Yuxin Cao, Youwei Shu, Derui Wang, Siji Chen

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 20 functions out of this paper's own repositories and ran 19 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
tdano1/eg-on-smoothing canonical 19 of 20
FunctionStatusWhere it lives
Int_for_gamma Ran tdano1/eg-on-smoothing/distribution.py
pointer only (licence: NONE) · get_code("ecc60901ba340fb0")
T Ran tdano1/eg-on-smoothing/th_heuristic.py
pointer only (licence: NONE) · get_code("46f3c7ec743814b7")
confidence_bound Ran tdano1/eg-on-smoothing/smooth.py
pointer only (licence: NONE) · get_code("2f3ad7fc799ce527")
confint Ran tdano1/eg-on-smoothing/distribution.py
pointer only (licence: NONE) · get_code("9a960e95cfbfcce0")
conv3x3 Ran tdano1/eg-on-smoothing/archs/cifar_resnet.py
pointer only (licence: NONE) · get_code("fac5364e2f53c6db")
conv3x3 Ran tdano1/eg-on-smoothing/archs/wide_resnet_imagenet64.py
pointer only (licence: NONE) · get_code("00e569acd6b45ef0")
fast_beta Ran tdano1/eg-on-smoothing/algo/algo.py
pointer only (licence: NONE) · get_code("cc016e2d63decb65")
get_beta Ran tdano1/eg-on-smoothing/th_heuristic.py
pointer only (licence: NONE) · get_code("2a24fb5dc7d5592e")
get_beta2 Ran tdano1/eg-on-smoothing/th_heuristic.py
pointer only (licence: NONE) · get_code("fadbf4d5b828f448")
get_input_shape Ran tdano1/eg-on-smoothing/datasets.py
pointer only (licence: NONE) · get_code("42a362aee3014e4b")
get_num_classes Ran tdano1/eg-on-smoothing/datasets.py
pointer only (licence: NONE) · get_code("d11c42096411b81b")
lambertWlog Ran tdano1/eg-on-smoothing/utils.py
pointer only (licence: NONE) · get_code("15c496ab663c5ec4")
ln_exp_plus_exp Ran tdano1/eg-on-smoothing/algo/algo.py
pointer only (licence: NONE) · get_code("431e5dacd9481349")
read_orig_Rs Ran tdano1/eg-on-smoothing/utils.py
pointer only (licence: NONE) · get_code("27a5ef2a70140e8c")
read_pAs Ran tdano1/eg-on-smoothing/utils.py
pointer only (licence: NONE) · get_code("51b78b0b7766cfbe")
sum_exp_greater_than_one Ran tdano1/eg-on-smoothing/algo/algo.py
pointer only (licence: NONE) · get_code("781b5c0335c1dea9")
u2x Ran tdano1/eg-on-smoothing/distribution.py
pointer only (licence: NONE) · get_code("72bd6ca75caa17d8")
wide_resnet_imagenet64 Ran tdano1/eg-on-smoothing/archs/wide_resnet_imagenet64.py
pointer only (licence: NONE) · get_code("36e1813d89519d14")
wide_resnet_imagenet64_1000class Ran tdano1/eg-on-smoothing/archs/wide_resnet_imagenet64.py
pointer only (licence: NONE) · get_code("75a74c940c2e498e")
get_dataset Not yet run tdano1/eg-on-smoothing/datasets.py
pointer only (licence: NONE) · get_code("ebd680ff3d6e442b")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Randomized Smoothing (RS) is currently a scalable certified defense method providing robustness certification against adversarial examples. Although significant progress has been achieved in providing defenses against ℓ p adversaries, the interaction between the smoothing distribution and the robustness certification still remains vague. In this work, we comprehensively study the effect of two families of distributions, named Exponential Standard Gaussian (ESG) and Exponential General Gaussian (EGG) distributions, on Randomized Smoothing and Double Sampling Randomized Smoothing (DSRS). We derive an analytic formula for ESG's certified radius, which converges to the origin formula of RS as the dimension d increases. Additionally, we prove that EGG can provide tighter constant factors than DSRS in providing Ω( √ d) lower bounds of ℓ 2 certified radius, and thus further addresses the curse of dimensionality in RS. Our experiments on realworld datasets confirm our theoretical analysis of the ESG distributions, that they provide almost the same certification under different exponents η for both RS and DSRS. In addition, EGG brings a significant improvement to the DSRS certification, but the mechanism can be different when the classifier properties are different. Compared to the primitive DSRS, the increase in certified accuracy provided by EGG is prominent, up to 6.4% on Im-ageNet. Our code is available at https://gi thub.com/tdano1/eg-on-smoothing.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.02309")
get_code_for_paper("2406.02309")
have("2406.02309")

Connect an agent — have() is free.