Inverse optimal control for modeling virus mutations in SARS-CoV-2

Abstract

Inverse Optimal Control (IOC) deals with the problem of recovering an unknown cost function in a Markov decision process from expert demonstrations acting optimally. In this thesis we apply IOC to SARS-CoV-2 data. For our application we use the (mutated) sequences found in SARS-CoV-2 data as the expert demonstrations. We present a way to learn useful state representations for this data, and successfully apply IOC on a special class of Markov decision processes which allow for an efficient computation of the value and cost functions of the states, and informative 2D representations of the state.