Published March 27, 2023 | Version v2
Software Open

Towards Global Neural Network Abstractions with Locally-Exact Reconstruction: Data, Networks and Code

Authors/Creators

  • 1. The University of Manchester
  • 2. Federal University of Amazonas

Description

Data, neural networks and code of the experiments in the paper "Towards Global Neural Network Abstractions with Locally-Exact Reconstruction". A stand-alone Python library to compute GINNACER is included, as well as state-of-the-art global and local abstraction alternatives.

Files

ginnacer_experiment_package_v2.zip

Files (58.5 MB)

Name Size Download all
md5:1eba881851c0acc97776c159e812c9ab
58.5 MB Preview Download

Additional details

Related works

Is documented by
Journal article: arxiv.org/abs/2210.12054 (Handle)

Funding

UK Research and Innovation
EnnCore: End-to-End Conceptual Guarding of Neural Architectures EP/T026995/1