Train, validate and test separately and how to retrain a neural network with another dataset without resetting its weights

Hi, I'm new to Matlab's Neural Network API and I'm having this problem:

I have to train a sequential set of data, let's say for example a gaussian curve. However, I can't find a way to train other sets of data without resetting this network. In other words, I want to update the weights using other shapes of gaussian curves.

The solution I'm looking for would have to:

• Train, validate and test separately and generating separate output files.
• Train the network without changing the order of the input
• Support "retraining" without resetting the weights.

If I wasn't clear in anything just tell me. I'm really needing an answer, this is part of my graduate project.

Thanks in advance.