How to use pytorch for deep learning
Terms¶
Epoch¶
Refers to each round of training with a given batch to minimize the loss as much as possible. As you increase the number of epochs, you should keep decreasing the loss.
Batch¶
Batch is the full set of training samples. Using all training samples for to compute the loss function is called batch gradient descent. Using samples, one-by-one, is called stochastic gradient descent. Using only a subset of training samples is called mini-batch gradient descent.