Learning the Wider Universe

October 2018

In areas as diverse as astronomy, cosmology, medicine, economics, and sociology, our ability to collect data is growing rapidly, providing important opportunities but also making it more challenging to thoroughly explore our data. Sparked by major new initiatives, scientists from different fields of research are employing "machine learning" techniques. Machine learning is a computer-based approach in which, rather than programming specific guidelines and procedures based on the science being explored, our programs teach the computer to learn in "non-parametric" ways: that is, in ways that do not rely on having detailed physical models. Rather than invoking scientific principles from biology or physics, for example, we write and use software that can identify trends, and also discover deviations from the "usual." As groups across the sciences make significant efforts to take advantage of machine learning and to develop it further, we plan to bring together scientists from across disciplines with the goal of learning from each other and exploring possible cross-disciplinary collaborations. We aim to make our research programs and those of our colleagues more effective by fostering collaborations designed to optimize opportunities for cross fertilization. Our interest in the possibilities that could come to fruition through an exploratory seminar is sparked by new data-intensive programs in astronomy and astrophysics, which have clear cross-disciplinary connections. In fact, the problems of "Big Data" are oft discussed and have been addressed in several Radcliffe Institute events and by a number of Radcliffe fellows.