nm-narasimha/iterative-proportional-fitting

Iterative Proportional fitting technique to create sample weights such that data is representative of the target dataset or distributions. In statistics, a sample statistic would be biased if sample is not representative of the "population". In ML/AB context, impact of an treatment in an experiment (treatment vs control) or impact of an attribute in an "explainable" model would be biased/incorrect if the distribution of the sample on features is not accurate.

Jupyter Notebook