Abstract: In inverse optimal transport (IOT) one wishes to identify unknown cost criteria from noisy or partial observations of optimal matchings. Consider for example commodity data; assuming that this data corresponds to the optimal match of buyers and vendors, can we identify the underlying cost?
In this talk I will discuss a systematic approach to infer unknown costs from noisy observations of optimal transportation plans, which is rooted in the Bayesian framework for inverse problems. The proposed methodologies can be used for various applications of OT problems in economics, transportation research and logistics, but I will illustrate them with an example from migration flows.
Joint work with Andrew Stuart (Caltech)