Summarizing and Analyzing Data using Optimal Transport
Now that optimal transport algorithms are reaching new levels of sophistication, we can turn our attention to developing mature applications of transport in machine learning and statistics. In this talk, I will demonstrate the breath of applications in which we can incorporate machinery from transport, as well as the computational techniques needed to make them work in practice. I will demonstrate how optimal transport can be used to represent complex data, fuse multiple learned models, and improve the performance and applicability of inference techniques. This work is a collaboration with the MIT Geometric Data Processing Group and colleagues at the MIT-IBM Watson AI Lab.