Learning and Decision-Making with Data: Optimal Formulations and Phase Transitions

arXiv:2109.06911v3 Announce Type: replace-cross
Abstract: We study the problem of designing optimal learning and decision-making formulations when only historical data is available. Prior work typically commits to a particular class of data-driven formulation and subsequently tries to establish out-of-sample performance guarantees. We take here the opposite approach. We define first a sensible yard stick with which to measure the quality of any data-driven formulation and subsequently seek to find an optimal such formulation. Informally, any data-driven formulation can be seen …

abstract arxiv class cs.lg data data-driven decision designing historical data making math.oc math.st performance prior sample stat.ml stat.th study transitions type work