I am using the Matlab neural network toolbox in order to train an ANN. From past experience, implementing cross validation when working with ML algorithms can help reduce the problem of overfitting, as well as allowing use of your entire available dataset without adding bias.
My question is; is there any advantage to implementing k-fold cross validation when using the NN toolbox, or are overfitting and bias mitigated by the implementation already (e.g. in the 'trainbr' mode)
Best Answer