What is the purpose of shuffling the validation set?

29 views (last 30 days)
Cynthia on 17 Jan 2020
Commented: Cynthia on 18 Jan 2020
What is the purpose of shuffling the validation set during training of an artificial neural network? I understand why this makes sense for the training set, so that each minibatch has a different composition every time, but doesn't the ANN evaluate the whole validation set every epoch?

Answers (1)

Greg Heath
Greg Heath on 18 Jan 2020
To impose and verify a consistent GENERALIZED path to convergence by avoiding repetitive anomalies.
Hope this helps
  1 Comment
Cynthia on 18 Jan 2020
I'm sorry, but I think I am asking a different question than the one you are answering. It is my fault for being unclear. At the most basic level, I need to understand which validation set it is that is being shuffled. Is it the one I specify in trainingOptions? Or is it a validation set that is invisible to me, made from the training set?

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!