Early Stopping
Always a good idea!
Stop training the model when accuracy on the validation set decreases.
Or train for a long time, but always keep track of the model snapshot that worked best on val.
Stop training the model when accuracy on the validation set decreases.
Or train for a long time, but always keep track of the model snapshot that worked best on val.