Inputs vs. Correct Output Data Scaling in Neural Networks

I am trying to use fitnet on a series of input data to validate with a series of correct outputs. While the input data magnitude sits around 10^-1, the correct output magnitude is much bigger, the data values are in the order of 10^3-10^4.

I know that scaling is a relevant step when using neural networks and I was wondering what could be the best approach in scaling inputs and outputs in order to better train a shallow network.