Authors: Utsav Sadana, Abhilash Chenreddy, Erick Delage, Alexandre Forel, Emma Frejinger, Thibaut Vidal

Published on: June 17, 2023

Impact Score: 8.15

Arxiv code: Arxiv:2306.10374

Summary

• What is new: This research surveys the burgeoning field of contextual optimization, compiling various models and methods from the realms of operations research and machine learning into a cohesive review.
• Why this is important: The challenge of making optimized decisions in uncertain circumstances by combining prediction algorithms and optimization techniques.
• What the research proposes: Identification of three main frameworks for learning policies from data in the context of single and two-stage stochastic programming problems, discussing their strengths and weaknesses.
• Results: The research organizes existing models and methods into a unified notation and taxonomy based on the identified frameworks, aiming to deepen understanding and encourage further developments in integrating machine learning with stochastic programming.

Technical Details

Technological frameworks used: Contextual optimization, single and two-stage stochastic programming

Models used: Data-driven optimization, prescriptive optimization, predictive stochastic programming, policy optimization, predict/estimate-then-optimize, decision-focused learning, end-to-end learning

Data used: Variety of datasets as applicable in operations research and machine learning communities for decision-making under uncertainty

Potential Impact

This insight could influence a wide range of sectors that rely on decision-making under uncertainty, including logistics, finance, healthcare, and tech companies focusing on predictive analytics and operational efficiency.

Want to implement this idea in a business?

We have generated a startup concept here: OptiML Solutions.