Hi Lorenzo,
From your description, here is a high-level overview of the steps that you can take to apply reinforcement of a regression neural network using MathWorks products, more specifically the "Reinforcement Learning Toolbox" :
1. Defining a custom environment using the "Reinforcement Learning Toolbox" involves defining the state space, action space, and reward function. Functions such as rlFunctionEnv or rlSimulinkEnv can further be utilized if you are integrating with Simulink.
2. Designing the neural network with the help of the "Deep Learning Toolbox", to perform regression tasks using continuous output layers. Functions such as dlnetwork can be utilized to build your network.
3. Initialize the RL agent with an appropriate agent from the "Reinforcement Learning Toolbox" and configure it using options objects such as rlDDPGAgentOptions, rlTD3AgentOptions, or rlSACAgentOptions.
4. Train your model using the train function and configure using the options to let the agent interact with the custom environment to collect experiences and update its neural network on a reward based metric.
5. Lastly, you can evaluate the performance of the neural network on the regression task using the sim function.
You can also refer to the below documentations to learn in-depth about the "Reinforcement Learning Toolbox" and "Deep Learning Toolbox" respectively:
Hope it helps!
