MLExpert logo


What is the difference between epoch, batch, and iteration?

  • Epoch: forward and backward pass over all examples in the dataset
  • Batch: number of examples processed in one forward and backward pass. Defined by the user.
  • Iterations: number of examples in the dataset / batch size

Let's walk through an example:

We have 8,000 training examples. One epoch would be passing through all those. Let's take a batch size of 8. That would make the total number of iterations:

Iterations=number of examplesbatch size=80008=1000\text{Iterations} = \frac{\text{number of examples}}{\text{batch size}} = \frac{8000}{8} = 1000

Copyright © 2021 MLExpert by Venelin Valkov. All rights reserved.