Syntactic vs Semantic Linear Abstraction and Refinement of Neural Networks

Publikation: Beitrag in Buch/Konferenzbericht/Sammelband/GutachtenBeitrag in KonferenzbandBeigetragenBegutachtung

Beitragende

Abstract

Abstraction is a key verification technique to improve scalability. However, its use for neural networks is so far extremely limited. Previous approaches for abstracting classification networks replace several neurons with one of them that is similar enough. We can classify the similarity as defined either syntactically (using quantities on the connections between neurons) or semantically (on the activation values of neurons for various inputs). Unfortunately, the previous approaches only achieve moderate reductions, when implemented at all. In this work, we provide a more flexible framework, where a neuron can be replaced with a linear combination of other neurons, improving the reduction. We apply this approach both on syntactic and semantic abstractions, and implement and evaluate them experimentally. Further, we introduce a refinement method for our abstractions, allowing for finding a better balance between reduction and precision.

Details

OriginalspracheEnglisch
TitelAutomated Technology for Verification and Analysis
Redakteure/-innenÉtienne André, Jun Sun
Herausgeber (Verlag)Springer, Cham
Seiten401-421
Seitenumfang21
ISBN (elektronisch)978-3-031-45329-8
ISBN (Print)978-3-031-45328-1
PublikationsstatusVeröffentlicht - 2023
Peer-Review-StatusJa

Publikationsreihe

ReiheLecture Notes in Computer Science, Volume 14215
ISSN0302-9743

Externe IDs

Scopus 85175948369
ORCID /0000-0002-3437-0240/work/165454710

Schlagworte

DFG-Fachsystematik nach Fachkollegium

Fächergruppen, Lehr- und Forschungsbereiche, Fachgebiete nach Destatis