PV1 ⊧ nnenum: The Neural Network Enumeration Toolgiven an unsafe set an a neural network, checks whether the set overlaps with the output of the network

Application domain/field

• Neural networks
• Neural network verification

Type of tool

Neural network verifier

Expected input

• Neural network
• Input set
• Unsafe set

?

Expected output

safe or unsafe. For neural networks this means that they check whether the output of the neural network (for a given input set) overlaps with the provided unsafe set. If they do not overlap, then the network is considered safe.

Internals

nnenum focuses on the verification of fully-connected, feedforward neural networks with ReLU activation functions.