Speaker: Joseph Daws, UTK
Title: A partial decryption of machine learning for mathematicians
Abstract: The recent explosion of research into machine learning, artificial intelligence and neural networks across many academic disciplines has led to a gigantic zoo of acronyms and concepts many of which are relatively small variations of a few core ideas. In this talk, we will discuss a slightly over-simplified view of machine learning which exposes natural connections between machine learning and mathematics. One of the main reasons behind the recent explosion of research is the astonishing success of neural networks used for solving complex problems such as image classification, playing games, and driving cares. These problems, however, are not so easily stated as classical mathematical problems of interest. Neural networks may also be applied to more typical mathematical problems. In particular, we will discuss using neural networks for function approximation, solving PDEs, and solving the shortest path problem.
Wednesday, February 12 at 3:35pm to 4:35pm
Ayres Hall, 113
1403 Circle Drive, Knoxville, TN 37996