Anna4142/Deep-Spiking-Q-Networks

Implementation of spiking DQN training using different conversion techniques and backpropagation with surrogate gradients employed on the OpenAIGym environments CartPole, MountainCar, and BreakOut. Accompanies the thesis "Deep Spiking Q Networks" submitted at the Technical University of Munich in January 2020.

Jupyter NotebookMIT