I am online training a RNN with fixed batch size k on a time series. Initially I train my model with n batches and a number of e epochs. When a new batch n+1 is available, I would like to update the weights, rather than retrain the model with all n+1 batches. My question is, should I train the model on the new batch with the same number of epochs e? Should the number of epochs effectively be 1? Is there any way to know what optimal number to choose? Obviously, the more epochs I use on the new batch, the more weight this new batch will have on my predictions. I would like to read more on this, so if anyone has references they are very welcome!
