Bremen 2017 – scientific program
AGPhil 5.3: Talk
Thursday, March 16, 2017, 18:30–19:00, GW2 B2900
Machine learning and realism — •Ryan Reece1 and Nico Formanek2 — 1University of California, Santa Cruz (UCSC) — 2Höchstleistungsrechenzentrum Stuttgart (HLRS)
Machine learning is bringing new methods for taming the torrents of big data facing today's scientific projects and businesses. Not only is machine learning bringing new ways of automating the processing of data, but also automating processes of making inferences on that data, including unsupervised classification, model selection, and model fitting (regression). We argue here that philosophers should be interested in these developments because they offer provocative ways of framing classic philosophical questions concerning the problem of induction, realism, and natural kinds, among others. Drawing on examples of uses of machine learning in particle physics analysis, we introduce and discuss the following questions: How does statistical hypothesis testing address the problem of induction? How can machine learning be used in statistical inference? Can the scientific method be automated? And if so, what does that imply about the objectivity of science? How is clustering related to natural kinds?
N.B.: This talk is complementary to the talk titled "The automated discovery of physical laws", which looks specifically at automated inference of physical laws.