"In online machine learning, the idea is that your model is only allowed to learn from one example at a time. This is contrast to batch learning where all the available data is processed at once."
Read it!To avoid responding with "that's what Andrew NG said" when asked about the reason behind choosing an 80% training and 20% validation split, consider this explanation.
Read it!