Jannis Kurtz

We show that for integer problems calculating optimal CEs is Sigma_2^p-hard and provide solution algorithms which are able to solve instances in small dimension.

Essentially a counterfactual question applied to optimization problems asks: what is the minimal change in the problem parameters which would lead to a different but favored optimal solution?

Counterfactual explanations (CE) recently received increasing attention as a tool to provide explanations for the decisions of optimization problems; see e.g. Kurtz et al. (2025).

In our new work we study #counterfactual #explanations for linear #integer #optimization problems:

optimization-online.org/2025/10/coun...

This is joint work with Felix Engelhardt, Ilker Birbil and Ted Ralphs

I think no real mathematician should ever be asked to come up with an application for their math. There should be a human right for this

@jannisku.bsky.social is giving a very interesting talk on explainable (integer) optimization at the @euroorml.bsky.social seminar, highlighting counterfactual explanations before diving on new work with Coherent Local Explanations for Mathematical Optimization (CLEMO): arxiv.org/abs/2502.04840

Meanwhile on LinkedIn: „Academic operations research is useless since papers don’t study what is useful for Amazon“

we need more anti capitalist operations research

Explainable Optimization Managed by XOpt Team @ UvA. A websited dedicated to collect and promote the recent research on explainable optimization (XOpt).

We recently set up a new webpage where we want to collect all advances in the newly emerging field of #explainable #optimization

www.expopt.org

Our webpage collects publications, events and software packages related to the topic. If you want your work to be added please use our contact page:

I will give a talk about our newest work in #Explainable #Optimization in the EURO OSS. You are welcome to join!

But to be fair, this type of buzz-word repetition you also find at business schools without LLM use 😆

Join us for the next Robust Optimization Webinar (ROW) this Friday, October 3, at 17:00 (CET) (please note the changed time).

Speaker: Paul Grigas (University of California, Berkeley)
Title: Parametric Optimization Beyond Discretization

More info on our webpage: sites.google.com/view/row-ser...

In collaboration with University of Amsterdam, Harvard Medical School and Massachusetts General Hospital we developed a heuristic algorithm for the k-adaptability approach applied to #robust treatment planning in radiation therapy.

Paper link: arxiv.org/abs/2508.07368

Robust Optimization Webinar The Robust Optimization Webinar (ROW) is a bi-weekly online seminar covering presentations of international researchers related to the topic of robust optimization. The seminar covers presentations on...

More information on our webpage: sites.google.com/view/row-ser...

The Robust Optimization Webinar (ROW) is back from the summer break! We will have our first seminar this Friday, September 5 at 15:00 (CET).

Speaker: Alexander Shapiro (Georgia Institute of Technology)

Title: Rectangularity and Duality of Distributionally Robust Markov Decision Processes

YouTube video by Policy-Relevant Science & Technology Munk Debate on Artificial Intelligence | Bengio & Tegmark vs. Mitchell & LeCun

But I mean the leading academics in the field are frequently debating the impact of AI (see e.g. www.youtube.com/watch?v=144u...). We can discuss how realistic or smart some arguments are but at least there is a discussion. Never saw anything coming close to this in OR.

Very true, although at least in AI research people are more aware of societal/political consequences of their works. In OR people had some discussions in the 80s and after the neoliberal agenda took over it feels there is no discussion about the inherent OR ideology anymore..

Our paper on #CounterfactualExplanations for #LinearOptimization Problems is now published and open accessible at European Journal of Operational Research:

doi.org/10.1016/j.ej...

The Robust Optimization Webinar (#ROW) will be back in September with Season 6. Meanwhile, to properly enjoy the summer, you can watch all presentations of our past speakers on our Youtube channel:

Youtube: youtube.com/@robustoptim...

Webpage: sites.google.com/view/row-ser...

The last #ROW of this season takes place Friday, June 27, at 15:00 (CET). We will be back with our 6th season beginning of September.

Speaker 1: Taylan Kargin (Caltech)

Speaker 2: Zihang Qiu (University of Amsterdam)

More information on our webpage: sites.google.com/view/row-ser...

Perfect showcase that for this type of problems using an LLM to search over the space of algorithms instead of smartly searching over the solution space is a complete waste of resources.

FICO Xpress Optimization Surpasses AlphaEvolve’s Achievements FICO’s global optimization solver outperforms Gemini AI-powered AlphaEvolve coding agent by improving solutions for complex, constrained math problems

Timo Berthold has a post called “FICO Xpress Optimization Surpasses AlphaEvolve’s Achievements” www.fico.com/blogs/best-g... #orms

4th AI & Mathematics in NL workshop in Tilburg.

Many cool presentations: aimath.nl/index.php/20...

And great people:

Join us for the next Robust Optimization Webinar (#ROW), this Friday, June 13, at 15:00 (CET):

Speaker: Ian Yihang Zhu (NUS Business School)

Title: Network Flow Models for Robust Binary Optimization with Selective Adaptability

More information on our webpage: sites.google.com/view/row-ser...

Ok these algorithm abbreviations are getting out of hand 😂

Maybe we can use chocolate instead. Chocolate is quite mainstream.

Join us for the next Robust Optimization Webinar (#ROW), this Friday, May 30, at 15:00 (CET).

Speaker 1: Zhao Kang (Eindhoven University of Technology)

Speaker 2: Abbas Khademi (University of Tehran)

More information on our webpage: lnkd.in/eujnAkBA

A Comprehensive Evaluation of Contemporary ML-Based Solvers for Combinatorial Optimization Machine learning (ML) has demonstrated considerable potential in supporting model design and optimization for combinatorial optimization (CO) problems. However, much of the progress to date has been e...

#Machine-learning based algorithms are still not able to beat deterministic general purpose solvers for #combinatorial #optimization.

arxiv.org/abs/2505.16952

Join us for the next Robust Optimization Webinar (#ROW) this Friday, May 16, at 15:00 (CET).

Speaker: Bahar Taskesen (The University of Chicago Booth School of Business)

Title: Distributionally Robust Linear Quadratic Control

More information on our webpage: sites.google.com/view/row-ser...

Next Wednesday, May 14th, I will give a presentation in the Explainable AI Seminars @ Imperial:

xaiseminars.doc.ic.ac.uk

I will talk about our recent work on #robust #counterfactuals and counterfactuals for #linear #optimization problems:

doi.org/10.1287/ijoc...
arxiv.org/abs/2405.15431