Sufficiently Deep Supervised Learning in High-Dimensions & Fast Prediction, with Applications

点击翻译按钮获取中文摘要

The research is dedicated to the development of a novel method for a fast androbust prediction of a parameter that is essential to a system such that it isrelated to other random variable(s) (r.v.) that either affects the system orbeing affected by the system. Here, the term, "robust" implies as reliabilitywhich is desired within the on-going 3rd wave of the developmental stage ofMachine Learning. Let the system variable be denoted as X and the associatedobservable as Y . Either, or both these r.v.s could be higher-dimensional aswell rather than being just scalars. In this Ph.D, I will learn the functionalrelationship f(.) between X and Y i.e. Y = f(X) to predict the values of X atwhich a (noisy) test datum on Y is observed, where the prediction will be fastand reliable.