SAVER: A Toolbox for Sampling-Based, Probabilistic Verification of Neural Networks

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Sivaramakrishnan, Vignesh, Kalagarla, Krishna C., Devonport, Rosalyn, Pilipovsky, Joshua, Tsiotras, Panagiotis, Oishi, Meeko
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913596567453696
author Sivaramakrishnan, Vignesh
Kalagarla, Krishna C.
Devonport, Rosalyn
Pilipovsky, Joshua
Tsiotras, Panagiotis
Oishi, Meeko
author_facet Sivaramakrishnan, Vignesh
Kalagarla, Krishna C.
Devonport, Rosalyn
Pilipovsky, Joshua
Tsiotras, Panagiotis
Oishi, Meeko
contents We present a neural network verification toolbox to 1) assess the probability of satisfaction of a constraint, and 2) synthesize a set expansion factor to achieve the probability of satisfaction. Specifically, the tool box establishes with a user-specified level of confidence whether the output of the neural network for a given input distribution is likely to be contained within a given set. Should the tool determine that the given set cannot satisfy the likelihood constraint, the tool also implements an approach outlined in this paper to alter the constraint set to ensure that the user-defined satisfaction probability is achieved. The toolbox is comprised of sampling-based approaches which exploit the properties of signed distance function to define set containment.
format Preprint
id arxiv_https___arxiv_org_abs_2412_02940
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle SAVER: A Toolbox for Sampling-Based, Probabilistic Verification of Neural Networks
Sivaramakrishnan, Vignesh
Kalagarla, Krishna C.
Devonport, Rosalyn
Pilipovsky, Joshua
Tsiotras, Panagiotis
Oishi, Meeko
Machine Learning
Artificial Intelligence
We present a neural network verification toolbox to 1) assess the probability of satisfaction of a constraint, and 2) synthesize a set expansion factor to achieve the probability of satisfaction. Specifically, the tool box establishes with a user-specified level of confidence whether the output of the neural network for a given input distribution is likely to be contained within a given set. Should the tool determine that the given set cannot satisfy the likelihood constraint, the tool also implements an approach outlined in this paper to alter the constraint set to ensure that the user-defined satisfaction probability is achieved. The toolbox is comprised of sampling-based approaches which exploit the properties of signed distance function to define set containment.
title SAVER: A Toolbox for Sampling-Based, Probabilistic Verification of Neural Networks
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2412.02940