Are the epochs equivalent to the iterations? I'm testing a multilayer perceptron with both scikit-learn and keras using the tensorflow backend.
Keras: epochs=X
Scikit:  max_iter=X
The epochs of keras are the same as the iterations of scikit learn???
Thanks
 A: I have no experience with SciKit Learn, however, in deep learning terminology an "iteration" is a gradient update step, while an epoch is a pass over the entire dataset. For example, if I have 1000 data points and am using a batch size of 100, every 10 iterations is a new epoch. See Epoch vs iteration when training neural networks.
A: Although the other answer is correct for the standard terminology, unfortunately scikit-learn uses non-standard terms here, as can be seen from the source: sklearn.neural_network.MLP{Classifier,Regressor} use max_iters to refer to the number of epochs. The relevant lines look essentially like:
for it in range(self.max_iter):
    X, y = shuffle(X, y)
    for batch in gen_batches(n_samples, batch_size):
        # compute gradients on batch
        # take a step along that gradient

    # check stopping criterion

where the gen_batches helper splits up the whole dataset.
I sent a pull request clarifying this as a result of this question; the docs are now clearer, so thanks for asking!
