DiffCoALG: Differentiable Learning of Combinatorial Algorithms
Workshop @ NeurIPS 2025
Embedded Files
About this Workshop
About this Workshop
Combinatorial algorithms are fundamental across a wide range of domains, owing to their ability to model optimization and decision-making tasks under complex constraints. These algorithms underpin practical applications such as vehicle routing, network and chip design, clustering and information retrieval. Combinatorial problems are also prominent in various areas of machine learning such as natural language processing and robotics. Recent research has focused on leveraging neural networks to design novel combinatorial algorithms and to come up with techniques that allow seamless integration of classic combinatorial algorithms in differentiable neural network architectures. Developments in this field, commonly referred to as neural combinatorial optimization, have raised several fundamental questions that span both theory and practice.
In this workshop, we take a broad perspective on designing differentiable algorithms for combinatorial optimization. The goal of the workshop is to explore novel ideas in the design and applications of neural combinatorial optimization, as well as to improve our theoretical understanding of existing methods.
Call For Papers
Call For Papers
We would like to invite original contributions (especially early research work) on the following topics:
Neural combinatorial algorithms for classic problems (travelling salesperson, boolean satisfiability, graph partitioning, mixed integer programming, etc.) and their applications (chip design, vehicle routing, etc.)
Generalization capabilities of neural combinatorial algorithms, including size or OOD generalization.
Neural combinatorial algorithms with certificates (approximation guarantees, optimality certificates, etc.).
Architecture design for neural combinatorial optimization (LLMs, diffusion models, etc.).
Understanding the role of training paradigms in neural combinatorial optimization: Reinforcement Learning vs Supervised Learning vs Self-Supervised learning.
Learning + Optimization: Learning to Optimize, learning to branch, etc.
Differentiating through discrete algorithms and operations (e.g., blackbox differentiation, differentiable shortest path algorithms, etc.)
Neural Algorithmic Reasoning
Rounding and loss function design for neural combinatorial optimization.
Expressive power of neural network architectures for combinatorial problems.
Benchmarking existing methods for combinatorial optimization.
The above list of topics is not exhaustive. If you are working on topics related to combinatorial optimization and machine learning, please consider submitting to and attending our workshop!
Submissions in the form of extended abstracts must be at most 6 pages long (not including references and an unlimited number of pages for supplemental material, which reviewers are not required to take into account) and adhere to the NeurIPS format. We accept submissions of early stage work or of papers that are currently under review. Please note that, according to the NeurIPS guidelines, work that has been published already at a conference is not eligible for presentation at a workshop. Submissions should be anonymized. The workshop will not have formal proceedings, but authors of accepted abstracts can choose to have either a link to an arxiv version of their paper or a pdf published on the workshop webpage. If the authors give us an arxiv link, we will link it here from the list of accepted papers on this webpage.
Important dates:
Submission deadline: September 6, 2025, 23:59 GMT
Author notification: September 21, 2025, AOE
Camera-ready deadline: November 30, 2025, 23:59 GMT
Workshop date: December 6, 2025
Openreview link https://openreview.net/group?id=NeurIPS.cc/2025/Workshop/DiffCoALG.
Please submit papers before the deadline above.
Speakers/Panelists
Farinaz Koushanfar
(UCSD)
Soledad Villar
(John Hopkins University)
Matthias Niepert (University of Stuttgart)
Vikas Garg
(Aalto University)
Bistra Dilkina
(USC)
Michael Galkin
(Google)
Felix Petersen
(Stanford)
Panel Discussions
Panel Discussions
11:15 AM: Panel on Neural Algorithmic Reasoning with Yusu Wang (UCSD), Christopher Morris (RWTH Aachen University), Yu He (Stanford), and Matthias Niepert (University of Stuttgart).
Workshop Schedule and Plan
Workshop Schedule and Plan
The workshop will consist of four keynote talks, two poster sessions, and a panel discussion, with enough time for scientific discussions throughout a full-day schedule. Each talk will be roughly 45 mins including questions -- 35 mins for the talk and 10 mins for Q&A. We plan to record each of the talks, along with its live Q&A session. The introductory remarks and the panel discussion will be live.
Livestream and live QA link: https://neurips.cc/virtual/2025/loc/san-diego/workshop/109535
Below is the detailed schedule. The timezone is Pacific Time.
08:00 AM - 08:45 AM: Poster Setup and Viewing
08:45 AM - 09:00 AM: Opening Remarks
09:00 AM - 09:30 AM: Keynote Talk: Bistra Dilkina (USC)
09:30 AM - 10:00 AM: Keynote Talk: Farinaz Koushanfar (UCSD) - From Origin to Owner: Provenance and Watermarking for AI-Based Optimization Solutions
10:00 AM - 10:15 AM: Coffee Break + Discussions + Networking
10:15 AM - 10:30 AM: Contributed talk (oral): Learning with Local Search MCMC Layers
10:30 AM - 10:45 AM: Contributed talk (oral): Filter Equivariant Functions: A symmetric account of length-general extrapolation on lists
10:45 AM - 11:15 AM: Keynote Talk: Vikas Garg (Aalto University)
11:15 AM - 12:00 PM: Panel Discussion on Neural Algorithmic Reasoning + Q&A
12:00 PM - 01:30 PM: Poster Session - 1 + Lunch Break + Networking
01:30 PM - 02:10 PM: Keynote Talk: Michael Galkin (Google)
02:15 PM - 03:15 PM: Panel Discussion on "Neural Combinatorial Optimization: What Works, What Doesn’t, and What’s Next?" + Q&A
03:15 PM - 03:30 PM: Coffee Break + Discussions
03:30 PM - 03:45 PM: Contributed talk (oral): ARC: Leveraging Compositional Representations for Cross-Problem Learning on VRPs
03:45 PM - 04:00 PM: Contributed talk (oral): Unsupervised Learning of Local Updates for Maximum Independent Set in Dynamic Graphs
04:00 PM - 05:00 PM: Poster Session - 2
Workshop Organizers
Workshop Organizers
Indradyumna Roy (IIT Bombay)
Abir De (IIT Bombay)
Soumen Chakrabarti (IIT Bombay)
Web and Social Media Chair
Web and Social Media Chair
Pritish Chakraborty (IIT Bombay)
Contact-Email: diffcoalg [at] gmail [dot] com
Contact-Email: diffcoalg [at] gmail [dot] com
Page updated
Google Sites
Report abuse