The next step is to define the neural network control policy.
Currently, OVERTVerify supports
fully connected neural networks with ReLU activation functions.
The last layer may have a linear activation function.
The control policy should be prescribed as an nnet file.
To see the details of this representation, check out this repository.