Contextual Stochastic Optimization: From data to decisions (26w5614)

Organizers

Utsav Sadana (University of Montreal)

Erick Delage (HEC Montréal)

Bistra Dilkina (University of Southern California)

Angelos Georghiou (University of Cyprus)

Phebe Vayanos (University of Southern California)

Description

The Banff International Research Station will host the "Contextual Stochastic Optimization: From data to decisions" workshop in Banff from February 22 - 27, 2025.


This workshop focuses on the integration of ML and stochastic optimization to address large-scale decision-making problems under uncertainty. By bringing together leading researchers from Operations Research and Computer Science, the event will explore the field of contextual stochastic optimization, which merges advancements from both disciplines to incorporate real-time data into the decision-making process. The aim is to drive theoretical and algorithmic progress, advancing data-driven decision-making across various domains.


%Decision-making problems under uncertainty are pervasive across various critical domains, including energy management, supply chain logistics, finance, and healthcare. These challenges have traditionally been addressed through stochastic and robust optimization by using the historical data on the uncertain parameters that affect our decisions. A growing number of methods have been proposed to prescribe actions that account for side-information while making best use of available historical data about both the covariates and the dependent uncertain parameters.


Contextual optimization is a rapidly evolving field that can prescribe decisions based on the most recently updated information about the covariates. By integrating covariates observed before decisions are made, this approach can significantly improve the performance and reliability of decision-support systems. Our workshop is centered around three core themes: the potential of integrated models for improving decision-making, robust and risk-aware decision-making methods, and the application of contextual optimization in real-world scenarios. These discussions will be crucial in identifying and overcoming the challenges of implementing these advanced techniques in various domains. % The objective of this workshop is to foster collaborations between Operations Research and ML researchers in order to make the best use of data.


In addition to exploring new methodologies, the workshop will address key challenges such as the explainability and interpretability of decisions, data privacy, and the practical deployment of models using contextual information. While these challenges are well-recognized in the ML community, particularly with the rise of large language models, they remain underexplored in the Operations Research community. By fostering collaboration and knowledge exchange, the workshop aims to develop scalable models, promote better benchmarking practices, and expand the applicability of contextual optimization to a wider range of real-world problems.


The Banff International Research Station
for Mathematical Innovation and Discovery (BIRS) is a collaborative
Canada-US-Mexico venture that provides an environment for creative
interaction as well as the exchange of ideas, knowledge, and methods
within the Mathematical Sciences, with related disciplines and with
industry. The research station is located at The Banff Centre in
Alberta and is supported by Canada's Natural Science and Engineering
Research Council (NSERC), the U.S. National Science Foundation (NSF),
and Alberta's Advanced Education and Technology.