Omar Mouchtaki is a PhD student at Columbia University in the Decision, Risk and Operations department. Prior to joining Columbia he earned a BS and MS in Applied Mathematics and Computer Science from Ecole Polytechnique (Paris). His research aims at bridging the gap between the practice and theory of data-driven decision-making by developing methodological tools to unveil the value of data and to better leverage these data for central operational problems such as inventory management, pricing and assortment optimization. His research has been recognized by a first place in the RMP Jeff McGill Student Paper Award and a finalist position in the George Nicholson Student Paper Competition and in the APS Best Student Paper Award.<\/p>\n"}],"event_agenda":false,"event_photo_gallery":false,"event_presentations":false,"event_custom_heading":[{"event_custom_title":"Abstract","event_custom_details":"
Despite the wide availability of data,\u00a0when considering a decision-making problem of interest,\u00a0\u201crelevant\u201d data may be limited in practice due to the heterogeneity of market characteristics: datasets are usually constructed in certain spatio-temporal contexts and have to be used in other ones. For instance, the demand for a given product may vary significantly over time or in different geographic locations.\u00a0How should one leverage such data for decision-making and what performance can one expect as a function of the data at hand?<\/p>\n
In this work, we propose a framework to understand the interplay between relevance of past data and performance of data-driven algorithms across all sample sizes, small and large. We demonstrate this framework by\u00a0anchoring our analysis around\u00a0the\u00a0contextual Newsvendor problem in which the decision-maker observes past demands and associated\u00a0contexts,\u00a0and needs to make a decision in a new context. In our model,\u00a0closeness of contexts is indicative of closeness in distributions but the distribution of past or future demands is unknown to the decision-maker.\u00a0We\u00a0analyze the performance of data-driven algorithms through a notion of context-dependent worst-case expected regret.\u00a0Our analysis significantly departs from classical concentration-based arguments.\u00a0In particular, our central methodological contribution is to characterize, in an\u00a0exact\u00a0fashion, for any given configuration of contexts, the worst-case regret\u00a0of any policy belonging to a broad class that includes most common algorithms for the Newsvendor problem (Sample Average Approximation (SAA), k-NN, Kernel methods,\u2026). This result in turn allows us to unveil fundamental insights on the actual learning behavior of these central policies, and the economics of data sizes. In particular, we show that, for the Newsvendor problem, very few samples go a long way towards good operational decisions.<\/p>\n"}],"event_enquiry_details":{"event_enq_full_name":"","event_enq_department":"","event_enq_email":"","event_enq_telephone":"","event_enq_website":""}}]