In deep learning, the concept of validation loss is to ensure that the model being trained is not currently overfitting the data. Is there a similar concept of overfitting in deep q learning?

Given that I have a fixed number of experiences already in a replay buffer and I train a q network by sampling from this buffer, would computing the validation loss (separate from the experiences in the replay buffer) help me to decide whether I should stop training the network?

For example, If my validation loss increases even though my train loss continues to decrease, I should stop training the training. Does deep learning validation loss also apply in the deep q network case?

Just to clarify again, no experiences are collected during the training of the DQN.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.