Tutorial sobre decision-focused learning

Este tutorial se realizará el día miércoles 28 de mayo, a las 12 horas, en la Sala I10.

Se pueden inscribir en https://forms.gle/CKqjzvaZgee9LtvX9

Abstracts:

Decision-focused surrogate modeling for mixed-integer linear optimization, Prof. Qi Zhang

Mixed-integer optimization is at the core of many online decision-making systems that demand frequent updates of decisions in real time. However, due to their combinatorial nature, mixed integer optimization models can be difficult to solve, rendering them often unsuitable for time critical online applications. To address this challenge, we develop a data-driven approach for constructing surrogate optimization models for mixed-integer linear programs (MILPs) in the form of linear programs (LPs) that can be solved much more efficiently than the original MILPs. We train these surrogate LPs in a decision-focused manner such that for different model inputs, they achieve the same or close to the same optimal solutions as the original MILPs. One key advantage of the proposed method is that it allows the incorporation of all the original MILP’s linear constraints, which significantly increases the likelihood of obtaining feasible predicted solutions. Results from two comprehensive computational case studies indicate that this decision-focused surrogate modeling approach is highly data-efficient and provides very accurate predictions of the optimal solutions. In these examples, the resulting surrogate LPs outperform state-of-the-art neural-network-based optimization proxies.

A gentle introduction to decision-focused learning. Víctor Bucarey López

In this tutorial, we will introduce the decision-focused learning paradigm, which integrates predictive modeling and optimization in end-to-end training. We will review key techniques for embedding this paradigm into machine learning pipelines. Finally, we will discuss exact methods for solving regret-minimizing linear regression problems.