Configuring ANN on TIBCO Statistica

[bilgin eser]

I'm training networks that predict breast cancer molecular subtypes. I have a question about " training, test, and validation" sets. Our trials have 140 to 220 participants, and when I split data "70%, 15%, and 15%," reviewers of the manuscript argued that test sample size is not enough. What should I do When I split data, the "70% train, 30% test" test error increasing.