How to reverse/undo "normalize" function?

Hi.

I wanna ask you guys about undoing or reversing after using "normalize" function.

I used this function before using as input data of Neural Network.

But after learning ANN models, I have to reverse or undo to check the Model's accuracy.

is there any way to reverse or undo "normalize" function?

thank you for reading.

I hope you guys have a nice day.

[Abdallah Ghazi Faisal Zaid Alkilani]

Why not normalize your reference and then do the comparison?

Generally speaking, depending on how you use the normalize function, it might not be reversible.

Normalize your output and your reference the same way, then do your comparison for accuracy.

[Srivardhan Gadila]

@HYEON TEAK OH, What do you mean by model's accuracy? Can you provide more details on what is your input data, target data and Is your ANN used for Classification or Regression. By the term accuracy I'm assuming that your network is used for classification purpose.

Answer 1

[Srivardhan Gadila]

As @Abdallah Ghazi Faisal Zaid Alkilani already mentioned the normalization might not be reversible. You can normalize the test data using the same exact normalization used for training data.

In case of Neural networks I would suggest to use the inbuilt Normalization

1. In case of Multilayer Shallow Neural Networks you might use net.inputs{i}.processFcns
2. In case of Deep Neural Networks you might make use of 'Normalization' name-value pair argument for the Input Layers available in Deep Learning Toolbox