About epoch, iteration and BatchSize

The epoch, iteration, and batchsize are often seen in deep learning, following their own understanding of the three differences:

(1) BatchSize: Batch size. In the deep learning, SGD training is generally used, that is, each training batchsize a sample training in the training concentration;

(2) Iteration:1 a iteration equal to the use of batchsize samples training once;

(3) Epoch:1 epoch equals the training of all samples in the training set;

For example, the training set has 1000 samples, batchsize=10, then:

After training the entire sample set requires:

100 times iteration,1 Epoch.

About BatchSize can look here.