Neural Network Abstraction for Accelerating Verification

While abstraction is a classic tool of verification, it is not often in use for verification of neural networks. We introduce an abstraction framework applicable to feed-forward networks. For the particular case of ReLU, we can provide error bounds incurred by the abstraction. We show how the abstraction reduces the size of the network, while preserving its accuracy and how verification results on the abstract network can be transferred back to the original network.

Authors

• Stefanie Mühlberger

Keywords

No keywords are indexed for this paper.