My research focusses on the integration of machine learning and constrained optimisation. The aim is to make constraint solving more human-aware by learning from the daily operational environment and its users.
AAAI 2024: Paper Learning to Learn in Interactive Constraint Acquisition: combining ML with symbolic constraint learning, with Dimos Tsouros and Senne Berden PDFCode
C&OR journal on our Amazon Last Mile routing approach, an extension of our 'Learn-and-route' work: Probability estimation and structured output prediction for learning preferences in last mile deliveryPDF
DSAA 2024: Paper Electricity Price Forecasting based on Order Books: a differentiable optimization approach, nice novel application of DFL in the electricity domain, as part of a novel approach to predict order books and then solve the Euphemia balancing algorithm to get the expected price. Nice collaboration with Léonard Tschora, Erwan Pierre, Marc Plantevit and Céline Robardet PDF
Math+ TES 2023 summer school talk "From data to decisions: Combinatorial optimisation with learned inputs slides
JAIR journal Efficiently explaining CSPs with unsatisfiable subset optimization, the most detailed description of our optimal unsatisfiable subset work and how to use it for step-wise explanation sequences PDF
CP 2023:
Paper Simplifying Step-wise Explanation Sequences, to improve explainable constraint solving. PDFcode
Paper Guided Bottom-Up Interactive Constraint Acquisition, more scalable acquisition and with out-of-the-box solvers (in CPMpy of course). PDFcode
Tutorial Explainable Constraint Solving - A Hands-On Tutorial, slidescode
Workshop invited talk Integrating Constraint Programming with Machine Learning and Explanation: The Sudoku Assistant App, at the PTHG workshopSlides
Workshop talk Holy Grail 2.0: From Natural Language to Constraint Models with Dimos Tsouros, Hélène Verhaeghe and Serdar Kadıoğlu, at the PTHG workshop
Workshop talks Things we underestimated while developing the CPMpy constraint modelling library and Breaking Constraint Modelling Languages with Metamorphic Testing (extended abstract) at the ModRef workshopSlides
Constraints journal paper Learn and route: learning implicit preferences for vehicle routing with Rocs Canoy, Victur Bucarey and Jayanta Mandi PDF
Beste Demo Award! for our Sudoku AI Assistant that combines machine learning and constraint solving techniques, to cleverly assist in solving pen-and-paper sudoku's.
Invited talk at the Data Science platform of the University of Vienna: Learning from User and Environment in Combinatorial OptimisationSlides
ICML 2022: Paper Predict and Optimize: Through the Lens of Learning to Rank which shows that learning-to-rank losses and techniques can be re-purposed, and our related to existing losses, for decision-focussed learning! paper, slides
AI and 5G? Yes, nice multidisciplinary collaboration with Politecnico di Milano, TU Eindhoven and TU Delft: Baseband-Function Placement with Multi-Task Traffic Prediction for 5G Radio Access Networks in IEEE Trans. on Network and Service Mgmt (2022) PDF
JFPC/PFIA 2022: Honoured to give the JFPC invited talk on Learning and Reasoning with Constraint Solving and the role of CPMpy to enable it! Slides
Lecture on Perception- and Preference-based Constraint Solving at the super exciting masterclass on Bridging the Gap between Machine Learning and Optimization, organized by Adam Elmachtoub and Elias Khalis. Slides
Paper Model-Based Algorithm Configuration with Adaptive Capping and Prior Distributions where we explore learning priors to speed up algorithm configuration, for example for use in CPMpy. PDFCode
ECMLPKDD 2022: priviliged to be co-program chair! Turns out it involves lots of data crunching to do right, see our program chair scripts repository.
CP 2022
Paper Learning Constraint Programming Models from Data using Generate-and-Aggregate, the paper resulting from our participation to the constraint learning holy grail challenge last year. Using CPMpy of course. PDFCode (TBD)
Paper Learning MAX-SAT Models from Examples using Genetic Algorithms and Knowledge Compilation resulting from now-PhD Senne Berden's master thesis. PDFCode
IJCAI 2022
Tutorial on Using constraint solvers as an oracle, with CPMpy on Saturday (T25), website
Workshop on Data Science Meets Optimisation (DSO) on Sunday (W50), website
Also co-organizing the Doctoral Concortium for PhD students, more info
Online seminar at GERAD, HEC Montreal: "Learning and Reasoning with Constraint Solving" slides
CP 2021
Paper Data Driven VRP: A Neural Network Model to Learn Hidden Preferences for VRP, challenged by very limited data but nice results with domain-specific architectures and loss functions; with Jayanta Mandi, Rocs Canoy and Victor Bucarey PDFvideocode
We won the PTHG-21 Constraint Acquisition Challenge! Grammar-based constraint learning, great collaboration with Mohit Kumar and Samuel Kolb, paper in preparation challenge videochallenge report
Doctoral Program panelist on 'Presenting your work' video
Position paper at the ModRef workshop on CCPCPPAPI, the need for a Common CP C++ API
Paper Contrastive Losses and Solution Caching for Predict-and-Optimize, scales very well thanks to novel caching, with Maxime Mulamba, Jayanta Mandi, Michelangelo Diligenti, Michele Lombardi and Victor Bucarey PDFvideocode
Paper Efficiently Explaining CSPs with Unsatisfiable Subset Optimization optimal MUSs, with meta-constraints and incrementality with Emilio Gamba and Bart Bogaerts PDFvideo
AAAI 2021: Knowledge Refactoring for Inductive Program Synthesis, combining CP and program induction, nice collaboration with Sebastijan Dumancic (KU Leuven) and Andrew Cropper (Oxford) PDF
Leuven.AI JRE Keynote "Tips for a succesful PhD, and how to win an award with it" watch the video!
NeurIPS 2020, super proud to have our paper Interior Point Solving for LP-based prediction+optimisation with my PhD student Jayanta Mandi accepted! PDFsummary videocode
DSAA 2020: accepted paper on Probability of Default Estimation with a Reject Option with PhD students Lize Coenen and Ahmed KA Abdullah, DOI
CPAIOR 2020: wonderful paper on Hybrid Classification and Reasoning for Image-based Constraint Solving, a nice collaboration with my PhD students Maxime Mulamba, Jayanta Mandi and Rocs Canoy PDFcode
EJOR journal paper Using Shared Sell-through Data to Forecast Wholesaler Demand in Multi-echelon Supply Chains with Jente van Belle and colleague Wouter Verbeke DOI
Zorgt Waze voor meer sluipverkeer? Men Universiteit van Vlaanderen filmpje met Waze als voorbeeld van de voordelen en valkuilen van artificiele intelligentie