Deep Learning
What is an epoch?
A full training pass over the entire dataset such that each example has been seen once.
Card 1 of 259. Answer: A full training pass over the entire dataset such that each example has been seen once.
A full training pass over the entire dataset such that each example has been seen once.
Card 1 of 259. Answer: A full training pass over the entire dataset such that each example has been seen once.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
An epoch is one full pass of the training algorithm over the entire training dataset, so that every example has been used once to update the model.
Epochs are split into batches. With 10,000 examples and a batch size of 100, one epoch is 100 iterations, meaning 100 gradient updates. Batch size sets how many examples go into each update; the number of epochs sets how many times the model sees the whole dataset.
Training usually runs for many epochs, since one pass is rarely enough to converge. Too few epochs underfit and too many overfit, so the number is normally chosen with early stopping: watch the validation loss after each epoch and stop when it stops improving. Data is typically reshuffled at the start of every epoch.