About this document
Unsupervised Verification of Autoencoders by Sakhawat is a document available to read on EtoBox.
This document discusses unsupervised neural network verification. It introduces challenges in defining unsupervised adversarial attacks and encoding autoencoders for SMT solvers. It also discusses contributions in basing verification on loss functions and defining unsupervised attacks. It then discusses challenges in verifying neural networks and training robust autoencoders, and contributions in decreasing subfunctions for faster verification and more robustness.
- Author
- Sakhawat
- Language
- EN