[P] Improving mulitclass classification accuracy with Jain's Fairness Index

This is a light implementation of the idea in the paper [Leveraging Uncertainties in Softmax Decision-Making Models for Low-Power IoT Devices](https://www.mdpi.com/1424-8220/20/16/4603/htm). Instead of finding uncertainties I have added Jain's Fairness Index as a addition to the loss function.

Gist: https://gist.github.com/Gananath/8d167384da7d3bc078650c73fab1a8dd

accuracy classification fairness index machinelearning