ML — Data Standardization vs Normalization

Both Normalization and Standardization are preprocessing steps we take to:

1. Reduce the size of the data. As we process data, lots of transformations are applied producing really large numbers by a series of multiplications and other operations.

If you do not reduce the size, your computer may not have enough memory to process the data and/or it could take much longer to process it.

2. Some algorithms are really sensitive to non-normalized data, so, please make sure to always normalize data.

Should I choose Normalization or Standardization?

Normalization is useful when there are no outliers.

Standardization should be used in cases where the data follows a Gaussian (Normal) distribution.

Standardization does not get affected by outliers because there is no predefined range of transformed features.

Hands-on

For hands-on, please check the following notebook availabe at Github:

https://github.com/saurater/deeplearningwithsamfaraday/blob/main/Data_Normalization_and_Standardization.ipynb