Training means providing standard input along with expected (correct) output.

Tested data set: giving examples as input to trained classifier to produce output to check accuracy of selected technique.

Thank your Sukhwinder Singh Sran for your reply.

When you are working with supervised algorithms ( in case ANN ) you need to correct weights which represent knowledge. these weight must be built based on data ( it the train data ). after the training ends we need to verify the ability of the algorithms on generalization so we used ( test data set).