This question already has an answer here:

I know that bigger batch size gives more accurate results, but I'm not sure which batch size is ideal given the following cases:

  1. Training on 65000 examples and validating on 13000 examples
  2. Training on 50000 examples and validating on 25000 examples
  3. Training on 71000 examples and validating on 4200 examples
  4. Training on 61000 examples and validating on 14000 examples
  5. Training on 56000 examples and validating on 19000 examples

How big should the minibatch size be and why?

System specs: 128GB RAM


marked as duplicate by Franck Dernoncourt, Michael Chernick, mdewey, jbowman, whuber Feb 12 '18 at 21:30

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • $\begingroup$ You should minimal context to the question: what you learn, how, that it is gradient descent... $\endgroup$ – Benoit Sanchez Feb 11 '18 at 18:54