A Combined Reward-Penalty Loss Function based Extreme Learning Machine for binary classification

In this paper, we have introduced a novel combined reward cum penalty loss function to measure the empirical risk in Extreme Learning Machine. The proposed Reward cum Penalty loss function based Extreme Learning Machine (RP-ELM) penalizes those data points which do not lie on the desired location and assigns reward for those data points which lie on the desired location. Unlike OP-ELM, the RP-ELM can utilize the full information contained in the training set which makes it to own better generalization ability.