Data pre - processing for artificial Neural Network

Hi everyone, I have a question. I am using the advance scripts made by Matlab for the creation of neural networks (by neural network tool).

In input to the network I gave power values ​​that had been statistically Normalized.

Once I finished training the network I realized that there is already a pre - processing function within the script that allows you to Normalize the data using the maximum and minimum values.

What I now ask myself is, are the results I got from the network, having both standardized and normalized max-min data correct? Or is it nonsense?

[Giuseppe D'Amico]

can anyone answer me?

Answer 1

[Tarunbir Gambhir]

Min-Max Normalization is usually done when the data has varying scales and the training model does not make any assumptions about the distribution of data. Like Artificial Neural Network, or K-nearest neighbours.

Standardization assumes that the data has a Gaussian distribution, and therefore is generally employed when the data has varying scales and the training algorithm assumes that the data follows a Gaussian distribution. Like linear regression, logistic regression, or linear discriminant analysis.

If your data does follow a Gaussian distribution, performing standardization on top of min-max normalization should give you the desired results despite being an extra step.