Extending Neural Network Verification to a Larger Family of Piece-wise Linear Activation Functions

László Antal; Hana Masara; Erika Ábrahám

doi:10.4204/EPTCS.395.4

Extending Neural Network Verification to a Larger Family of Piece-wise Linear Activation Functions

Machine Learning 2023-11-21 v1 Artificial Intelligence Logic in Computer Science

Authors: László Antal , Hana Masara , Erika Ábrahám

View on arXiv ↗ PDF ↗ DOI ↗

Abstract

In this paper, we extend an available neural network verification technique to support a wider class of piece-wise linear activation functions. Furthermore, we extend the algorithms, which provide in their original form exact respectively over-approximative results for bounded input sets represented as start sets, to allow also unbounded input set. We implemented our algorithms and demonstrated their effectiveness in some case studies.

Keywords

neural network verification piece-wise linear activation over-approximation unbounded input exact analysis start sets

Cite

@article{arxiv.2311.10780,
  title  = {Extending Neural Network Verification to a Larger Family of Piece-wise Linear Activation Functions},
  author = {László Antal and Hana Masara and Erika Ábrahám},
  journal= {arXiv preprint arXiv:2311.10780},
  year   = {2023}
}

Comments

In Proceedings FMAS 2023, arXiv:2311.08987

Extending Neural Network Verification to a Larger Family of Piece-wise Linear Activation Functions

Abstract

Keywords

Cite

Comments

Related papers