Is it alright to apply different epoch numbers to train/test a ANN method ( i.e. set number of iterations/epoch for training mode, and then set another different number of epoch for testing mode) ?
There is no "number of epoch for testing mode". You need just check all test data outputs and compare them with real outputs after training. If result is fine - your training was successfully and you can use your network for unknown data, if not - you need make you learning more stable or make changes in learning process.