Fusion with Optimal Transport
Justin Solomon (Massachusetts Institute of Technology)
Many problems in learning and statistics involve fusing potentially conflicting signals into a single coherent observation about a system or environment. Optimal transport provides a valuable language for posing and solving such fusion problems in a mathematically-justified and efficient framework. In this talk, I will summarize efforts in our group on developing efficient algorithms for model fusion drawing from state-of-the-art algorithms for computational transport.