From Bayes to Machine Learning

Roberto Trotta, Imperial College London and Björn Malte Schäfer, Heidelberg University

Abstract:

Statistical inference and inductive reasoning are at the core of scientific progress: this course will introduce concepts of statistics and Bayesian methods for inference problems in science, both for deriving constraints on model parameters and for establishing the evidence in favour (and against) model classes as possible explanations for Nature, and the relative computational tools. At the end, basic concepts in machine learning will be introduced, as a novel and powerful way for finding patterns in large and complex data. We discuss fundamental theoretical assumptions such as objectivity and the choice of a prior, as well as information theoretical ideas.