Can we train with same data?

My question is after finish training the model, can we use back the same data shuffel it and use it to retrain model?

During training, you usually use all your training data, shuffled already. One epoch is one pass through all your training data. You usually train for more epochs, maybe 10 or 100 or more, which means you train your model on the same training data multiple times.

Usually, you have a validation set as well. After hyperparameter tuning, you can combine your training set and your validation set and train the model a final time before examining the model on the test set.

1 Like