Hi Sung,
Is your cumulative reward increasing? In the RL context, loss is actually not a very good measure of performance. There are actually certain instances where loss can increase and this corresponds to better performance.
Hi Sung,
Is your cumulative reward increasing? In the RL context, loss is actually not a very good measure of performance. There are actually certain instances where loss can increase and this corresponds to better performance.
PhD. Interests include Deep (Reinforcement) Learning, Computational Neuroscience, and Phenomenology.
PhD. Interests include Deep (Reinforcement) Learning, Computational Neuroscience, and Phenomenology.