Multiblock ADMM in Machine Learning

In this article we consider using modern fast alternation direction optimization methods for cancer diagnostics. During last years the number of people that suffer from cancer significantly increased. The aim of this work was to make research and to determine the most important factors that influence durability of life after surgery and to predict this durability. We collect data from a group of people with melanoma, define 36 characteristics that can influence durability of life and with help of LASSO problem and multiblock ADMM with Nesterov acceleration we determine the most important characteristics, f in d life expectancy and calculate the difference between real life durability and predicted values. Multiblock ADMM with Nesterov acceleration is an optimization method, based on combination of two approaches - splitting of original optimiza problem into N subproblems and finding Nesterov acceleration step on each iteration.