I have a beginner experience in using the Matlab, Neural Network Toolbox. My question is: I have some N x 22 feature matrix for training data. Where N is very large. In addition, there are three class labels (targets) corresponding to each of the N samples. I am wondering whether I can train a neural network (in general a classifier) using the training data only. So that I can use this trained classifier to test my test samples one-by-one during the testing phase.
To be precise, I want to save my trained classifier (neural network or SVM or k-NN). Later it should provide the predicted class label/target given a test sample (one-by-one in a for-loop). Also, I came across dividing the data (<http://www.mathworks.com/help/nnet/ug/divide-data-for-optimal-neural-network-training.html>) and similar community Q/A but I am still skeptical in order to use this for a single test sample.
Thank you for your time and any help is much appreciated.