0

I want to build an agent for binary classification. I have a large dataset with two label (0 and 1). I want to build an agent to predict labels. I build a deep model and now I want to build an agent. I use keras-rl2. but there is a problem: for dqn agent, the fit function has an env argument. I don't know how can I define my problem environment for that. note that my problem has a similarity function that optimize weights for each feature. the agent can also select the best weight for each feature. but the problem is that I don't know how can I define my environment.

sdbvuf sbjdsfdib
  • 1
  • 1
  • 1
    This sounds like the standard setup for supervised learning, not reinforcement learning. – Dave Jun 17 '21 at 09:41
  • 3
    There is not usually a good reason to apply reinforcement learning to a supervised learning problem. It will be slower to train, and is likely to produce a model that is less accurate at the end. *Sometimes* there are reasons why it is worthwhile. Could you clarify why you want to apply reinforcement learning in your case? – Neil Slater Jun 17 '21 at 16:27
  • I have an agent that each time sees an instance, it should select an instance in memory that has the most similarity with that. I just want to train the agent to select the best instance. @NeilSlater – sdbvuf sbjdsfdib Jun 19 '21 at 11:55
  • That doesn't seem like a reinforcement learning problem. Unless there is a consequence to the agent's situation other than being right or wrong, then there is normally no reason to use RL fo rthis kind of identification or classification task. – Neil Slater Jun 19 '21 at 14:14

0 Answers0