Dartmouth Events

Abstract:  Machine learning is all the rage these days!   Applications range from predicting consumer spending patterns, to understanding the contents of books and movies, from anticipating major moves in the stock market to driving your car or flying your plane, and in the process handily beating the world’s smartest people at games like chess and Go.  A 2015 MIT Sloan Management Review reported that 40 percent of the companies surveyed were struggling to find and retain their data analytics talent, and by 2018 there would be a shortfall of 181,000 data scientists.  Clearly, a huge market with seemingly endless applications exists, but what about the world of science?  Could machine learning be of value in understanding some of the data that we currently have, or will collect in the near future?  In this talk, I will make the point that in certain cases, machine learning is the only viable alternative to analyzing the sorts of data volumes that we can soon expect to collect. I will show examples of a neural-network-based reconstruction of a recent data set from Van Allen Probes and suggest a few ideas for incorporating these new tools into our existing paradigm of data analysis.

Jointly sponsored by Physics and Astronomy and The Society of Fellows at Dartmouth College