I am new to Reinforcement Learning. I have been working on a problem using Deep Deterministic Policy Gradient (DDPG).
I would like to know if it is possible to apply this algorithm to an optimization problem where I have both time-independent and time-dependent manipulated variables. This would mean that the Actor-network would output two different actions (2-dimensional vector). Considering that an episode is made of X number of timesteps, one action would obtain the same value throughout the entire episode and the second action would be changing at every timestep. The idea is to call the exploration and exploitation at different frequencies for each action.
When I run my code in this way, I get reasonable results, however, I am not sure if this is correct.
