neural

We extend our previous work on verifying agent-environment systems with an agent executing a ReLU-based controller to handle RNNs.

We develop a novel model for studying agent-environment systems, where the agents are implemented via neural networks.