36th Meet up: GPU accelerated κν algorithm for regression problems

The kappa-algorithm has shown significant robustness and promise within the machine learning domain, however one significant drawback is it's computational complexity. By it's very design the algorithm is O(n^2). While this may not prove to be a problem during inference time, it may be a problem during the training period for finding the optimal parameters. This hurdle has been seen in the industrial setting, which led to development of heurestic methods for determinig the optimal tuning paremerts. We take a different approach and adapt the algorithm for use in modern graphics processing units specifically using Nvidia's CUDA architecture.