I am on working on a problem that involves the use of Neural Network and or Reinforcement learning for Real-time optimization of a system/Plant. There are Four inputs and One output. The Task is to optimize the output given the constraints. Constraints are not on the output or input, they are some other parameters.
I have collected DOE data using simulation and have made a Neural network which is working great and is predicting the output very well(meeting my requirements). I have been able to do offline constrained optimization using the neural network I made and using genetic algorithm function in Matlab.
For the above-mentioned work Neural network was trained separately using the previous data collected and then optimized. I want to do everything online meaning, Data collection and optimization using some rewards as done in Reinforcmeent learning . Can somebody please help me steer me in the right direction. I can provide information if needed.