Batch vs Online Learning

Batch learning

• Training is performed with all the data

• It is performed offline

• It requires more resources and time (every time the training is performed)

Online learning

• Training is performed on mini-batches

• It is offline, as well

• It updates the model (not overwritten) - so, it requires fewer resources.

• It is important to properly set the regularization parameter and check the wellness of all the data fed into the model.