Overview

Bridging Machine Learning (ML) & Combinatorial Optimization (CO): from ML for CO to CO for ML. Fostering collaboration for breakthroughs, foundation models & real-world impact.

Combinatorial optimization (CO) addresses diverse real-world challenges spanning vehicle routing, chip design, smart city planning, and drug discovery. However, it has been a long-standing challenge to develop effective heuristics in solving different CO problems due to their discrete and non-convex nature, which typically demands significant human effort. In recent years, AI has increasingly contributed to solving CO problems, bringing learning-based heuristics, reinforcement learning, and neural-guided search into the mix. This makes CO not only practical but also fertile ground for methodologically impactful research.

At the same time, many core AI challenges can be formulated as CO problems. Examples include (i) discrete sampling: LLM decoding and beam search, discrete diffusion models; (ii) neural‐network design: model pruning, model merging, neural architecture search; and (iii) matching/alignment tasks: object detection, entity matching, cross‐domain alignment. These are not edge cases but fundamental, high‐impact tasks at the heart of modern AI systems.

This workshop aims to bridge the gap between the CO and AI communities and encourage deeper collaboration. To highlight impact in both directions, we will feature invited speakers from two complementary tracks: (a) ML for CO Track, with researchers advancing and applying ML methods for CO; and (b) CO for ML Track, with leading researchers introducing powerful tools that can revolutionize AI applications, and particularly advanced search methodologies.

Scope and Goals

This workshop explores the growing intersection of machine learning (ML) and combinatorial optimization (CO), with a focus on both ML4CO and CO4ML. We welcome contributions across a wide range of topics within ML4CO, including but not limited to Neural-based CO Solvers, GPU-accelerated solvers, Foundation Models for CO, LLM for CO, and Learning-guided Optimization. On the CO4ML side, we invite work that applies combinatorial techniques to solve structured machine learning tasks, such as Neural-Network Design, Matching/Alignment tasks, and CO for Discrete Sampling.

By bringing together these communities, we aim to foster new collaborations and inspire innovative research directions at the intersection of ML and CO. By fostering shared understanding and encouraging contributions to common infrastructure—such as benchmarks, scalable tools, and system-level integrations—we hope to accelerate mutual progress and strengthen the foundation for future breakthroughs.

Speakers

Dimitri Bertsekas

Arizona State University

Dimitri Bertsekas is a distinguished computational scientist and Fulton Professor of Computational Decision Making at Arizona State University. Previously, he spent four decades at MIT (1979-2019), following earlier faculty positions at Stanford University and the University of Illinois, Urbana. He has authored twenty influential books and research monographs, including seminal works on dynamic programming, reinforcement learning, and convex optimization. He earned numerous prestigious awards, including the Khachiyan Prize for Life-Time Accomplishments in Optimization (2014).

Topic: Reinforcement Learning, Dynamic Programming, and Model Predictive Control.

Christopher Morris

RWTH Aachen University

Christopher Morris is an Assistant Professor at RWTH Aachen University, where he leads the Learning on Graphs (LoG) group and holds a DFG Emmy Noether Fellowship. His research focuses on graph representation learning and the intersection of machine learning with combinatorial optimization. Before RWTH, he was a postdoc at Mila - Quebec AI Institute and McGill University, after receiving his Ph.D. from TU Dortmund University.

Topic: The Connection between ILPs and MPNNs.

Yingqian Zhang

Eindhoven University of Technology

Yingqian Zhang is an Associate Professor in the Information Systems group at Eindhoven University of Technology (TU/e). She develops AI solutions for decision-making, focusing on combining machine learning and deep reinforcement learning with (trustworthy) data-driven optimization and socially aware algorithms. Her work is applied to real-world challenges in smart mobility, sustainable logistics, and the energy transition.

Topic: Application of GNNs and LLMs in Solving Real-World CO Problems.

Yuandong Tian

Meta GenAI

Yuandong Tian is a Research Scientist Director in Meta GenAI, leading a group for Llama reasoning. His research direction covers multiple aspects of decision making, including reinforcement learning, planning and efficiency, as well as theoretical understanding of LLMs. Prior to that, he worked in Google Self-driving Car team in 2013-2014 and received a Ph.D in Robotics Institute, Carnegie Mellon University in 2013.

Topic: Advancing LLMs and Reasoning with Optimization.

Yiming Yang

Carnegie Mellon University

Yiming Yang is a Full Professor at Carnegie Mellon University (CMU), affiliated with both the Language Technologies Institute and the Machine Learning Department. Her research interests span large language models, multimodal foundation models, and generative AI for science. She has published over 400 papers, with multiple best paper awards at KDIR, SIGKDD, IJCAI, SCAMC and NCIP.

Topic: ML-Guided Sampling Methods for Combinatorial Optimization.

Bistra Dilkina

University of Southern California

Bistra Dilkina is an Associate Professor of Computer Science at the University of Southern California. She is also the co-Director of the USC Center for AI in Society (CAIS), a joint effort between the USC Viterbi School of Engineering and the USC Suzanne Dworak-Peck School of Social Work. She received her PhD from Cornell University in 2012, and was a Post-Doctoral Associate at the Institute for Computational Sustainability until 2013. Her work spans discrete optimization, network design, stochastic optimization, and machine learning.

Topic: ML-Guided MIP solving via Contrastive Loss.

Financial Support

We plan to offer the following financial support: 1) Registration Fee Grants for individuals experiencing economic hardship and members of underrepresented groups, and 2) a Best Paper Award. We are pleased to announce that the following companies have agreed to offer financial sponsorship:

We plan to seek additional sponsorships from leading companies to further broaden the workshop's accessibility.

Call For Papers

For detailed submission guidelines and more information, please refer to the call for papers.

Workshop Schedule

Organizers

Sirui Li

Senior Research Engineer, Microsoft Research

Federico Berto

PhD Student, KAIST

Yining Ma

Postdoctoral Associate, MIT

Cathy Wu

Associate Professor, MIT

Ido Greenberg

Senior Research Scientist, NVIDIA

Eli Meirom

Senior Research Scientist, NVIDIA

Fei Liu

Postdoctoral Researcher, CityUHK

Chuanbo Hua

PhD Student, KAIST

Shengyu Feng

PhD Student, CMU

Weiwei Sun

PhD Student, CMU

Gal Chechik

Professor, Bar-Ilan University & Director of AI, NVIDIA

Jinkyoo Park

Founder & CEO, Omelet & Associate Professor, KAIST

Community Building

Collaboration across machine learning, operations research, and diverse ML applications is essential for the success of ML and CO. This workshop will serve as a platform to spark dialogue, inspire novel solutions, and drive future advances. We aim to build lasting connections among researchers, practitioners, and industry professionals. Designed as the start of an ongoing workshop series, we will foster a growing, supportive community with long-term engagement.

We have launched the AI4CO Community which includes an active Slack channel that already hosts 330+ members and is open to all interested in ongoing collaboration. Join our community to stay connected with the latest developments in the intersection of AI and combinatorial optimization.

Venue

111 Harbor Dr, San Diego, CA 92101, United States