Orientation of data set before training simple ANN's

Well, I am new to implementing ANN's and there is something that i want to know. It maybe a bit silly though.

I just wanted to know that if we have a simple data set say dependent only on a single variable x1, and if we want to train the ANN to predict the data set values (no new values, only the data set values needs to be predicted) will sorting or arranging the data set in some way beforehand make the ANN go to the optimum value quicker or it really doesn't matter?

If arranging does matter than how do we arrange multidimensional data and what is the intuition behind it being faster? Just like in Binary Search for a random data set we get a performance gain of O(n^2)-O(nlogn) over Linear Search similarly I want to know whether such performance gains are available in ANN's in any way?

Note: I am aware this is already taken care by Batch Learning so i am specifically talking about Example by Example learning. Also, I want to know if some arranged form of data set works well for certain forms of training e.g. Momentum method, Delta method.