Part of the task is to define our own reward function (in get_reward.py). What is its use? Can we directly apply some training algorithms after defining that? We have tried to apply the algorithms provided in CityLearn directly but encountered numerous problems.
Hi @andy8 thanks for your question. The default reward function that was provided by the organizers may or may not be one that maximizes the agent’s performance and the evaluation scores hence, the provision for the reward function to be modified.
There are examples here where other participants have been able to plug in the algorithms to the environment.