I was wondering, in a NN, i understand you can split the dataset using for example divederand or divideblock. But how do you "save" the test set from running when training ? Also i understand you can divde and hold out part of the dataset with for example c = cvpartition(n,'Holdout',p), but this only divides into two parts training and test set. I am new to ML, so this is all a bit confusing still i hope this makes sense to you. Also what is the difference between cross validation and holding out one part of the dataset?