David Wolpert (Los Alamos National Laboratory; SFI External Professor)
Abstract. Statistical inference — machine learning — is central to Science in that it is the foundation of data analysis. I argue here that its relationship with Science is both far broader and far deeper. I illustrate this breadth by showing how it can be applied to many numerical methods that lie at the heart of Science, e.g., genetic algorithms and Monte Carlo methods. I then illustrate that some of the deepest issues in Science, like the legitimacy of building models, are intrinsically issues in (Bayesian) statistics, and that how we do that statistics can have huge practical implications.