How can I get the inverse of a transformation applied on dataset in ml studio designer

Question

Hi everyone,
I am creating a ML pipeline in Azure ML Studio Designer. I'm having an issue where I need to perform data normalisation at one point in my pipeline and I need to then reverse that transformation later in my pipeline.
Is there a way for me to reverse this transformation through the designer, or get access to the parameters used in the transformation so I can apply the inverse manually in a python notebook?

Answer 1

When you apply a transformation (e.g., normalization), the transformation parameters (such as the mean and standard deviation for normalization) are crucial for reversing the transformation.

Use a custom Python script within your pipeline to access and store these parameters after applying the transformation. The same parameters can later be used to reverse the normalization.

In the later stages of your pipeline, you can create a Python notebook step where you apply the inverse of the transformation.

For example, if you're using Min-Max Normalization, you would reverse it using the formula:

original value=normalized value×(max−min)+min

Similarly, for Z-score normalization, use:

original value=(normalized value×standard deviation)+mean

# Example for reversing z-score normalization
def inverse_normalization(normalized_data, mean, std_dev):
    return (normalized_data * std_dev) + mean

# Apply inverse to the dataset
original_data = inverse_normalization(normalized_dataset, mean_value, std_dev_value)