Advertisement · 728 × 90

Posts by Research Campus MODAL

SCIP Optimization Suite 10.0: Exact Solving, Better Decompositions, and a More Productive Ecosystem TL;DR: SCIP Optimization Suite 10.0 brings a numerically exact solving mode for rational MILPs, noticeable performance gains for MILP/MINLP, stronger presolving and symmetry handling, better heuristics and conflict analysis, IIS detection, and major updates to GCG, PaPILO, PySCIPOpt, and MIP-DD.

SCIP Optimization Suite 10.0 is out 🎉

→ Numerically exact solving mode for rational MILPs
→ 9-20% faster on hard MINLPs
→ IIS detection for debugging infeasible models
→ Better symmetry handling
... and many more

For full details see:
www.pokutta.com/blog/resear...

#optimization

4 months ago 2 2 0 0
High Purcell enhancement in all-TMDC nanobeam resonator designs with active monolayers for nanolasers We propose a nanobeam resonator incorporating an active monolayer, designed to achieve a high Purcell enhancement. The resonator is fully composed of transition-metal-dichalcogenide materials and inte...

All-TMDC nanobeam resonator designs for nanolasers: article published in @physrevb.bsky.social
dx.doi.org/10.1103/nxh9...
#NumericalOptimization #photonics #numerics @zuseinstitute.bsky.social @felixbinkowski.bsky.social @mathplusberlin.bsky.social @fc-modal.bsky.social

4 months ago 1 2 0 0
By default, SCIP comes with a bouquet of different plugins for solving MIPs and MINLPs. SCIP is currently one of the fastest non-commercial solvers for mixed integer programming (MIP) and mixed integer nonlinear programming (MINLP). If you are new to SCIP, want to dive in and don't know where to begin, then have a look at the documentation at https://www.scipopt.org

By default, SCIP comes with a bouquet of different plugins for solving MIPs and MINLPs. SCIP is currently one of the fastest non-commercial solvers for mixed integer programming (MIP) and mixed integer nonlinear programming (MINLP). If you are new to SCIP, want to dive in and don't know where to begin, then have a look at the documentation at https://www.scipopt.org

The #SCIP Optimization Suite 10.0.0 has been released!
Many new features, including exact MILP solving. Release report: optimization-online.org/2025/11/the-... Download this new major release at www.scipopt.org #Optimization mixed integer programming #MIP mixed integer nonlinear programming #MINLP

4 months ago 1 1 0 0
Physics-informed Bayesian optimization of expensive-to-evaluate black-box functions - IOPscience Physics-informed Bayesian optimization of expensive-to-evaluate black-box functions, Sekulic, Ivan, Schaible, Jonas, Müller, Gabriel, Plock, Matthias, Burger, Sven, Martínez-Lahuerta, Víctor José, Gaa...

Physics-informed Bayesian optimization of expensive-to-evaluate black-box functions - just accepted in @ioppublishing.bsky.social MACHINE LEARNING SCIENCE AND TECHNOLOGY dx.doi.org/10.1088/2632... @jcmwave.bsky.social, @zuseinstitute.bsky.social @mathplusberlin.bsky.social @fc-modal.bsky.social‬

5 months ago 2 3 0 0

Prof Tim Conrad on Mathematical Algorithms for Health Care Applications: Article (in German) @ WISTA-Magazin www.wista.de/aktuelles/ne... . Focus on mathematical methods for the analysis of large amounts of biomedical data at @zuseinstitute.bsky.social and within Research Campus MODAL #BMFTR #FUHUB

5 months ago 0 1 0 0
General assembly of Forschungscampus MODAL at FUHUB Berlin, November 2025

General assembly of Forschungscampus MODAL at FUHUB Berlin, November 2025

General assembly of Forschungscampus MODAL at #FUHUB Inspiring ideas, lively discussions, and shared visions. Thank you to all participants, both on-site and online, for your contributions and commitment! We look forward to the next steps—working together for practical research and innovation!

5 months ago 0 0 0 0
Luka Stärk and Carolin Scholl win the DISPLIB 2025 Competition for real-time train dispatching. Their custom branch-and-bound algorithm solved the difficult scheduling problems, many of them even to proven optimality. The innovative method has a parameterized complexity of O(4^k poly(n)), where k is the number of conflicts and n is the number of operations. The international competition was held in two phases, with problem instances published in October 2024 and February 2025, respectively. The winners were announced at the International Conference on Optimization and Decision Science (ODS 2025), which took place from September 1 to 4 in Milan. Luka Stärk will be researching train dispatching in the MobilityLab at the MODAL research campus starting in November. The photo shows Luka Stärk receiving the award, with the chairs of the Scientific and Organizing Committees of the Competition, Giorgio Sartor and Bjørnar Luteberget, on his left and right, respectively. Congratulations to Carolin and Luka!

Luka Stärk and Carolin Scholl win the DISPLIB 2025 Competition for real-time train dispatching. Their custom branch-and-bound algorithm solved the difficult scheduling problems, many of them even to proven optimality. The innovative method has a parameterized complexity of O(4^k poly(n)), where k is the number of conflicts and n is the number of operations. The international competition was held in two phases, with problem instances published in October 2024 and February 2025, respectively. The winners were announced at the International Conference on Optimization and Decision Science (ODS 2025), which took place from September 1 to 4 in Milan. Luka Stärk will be researching train dispatching in the MobilityLab at the MODAL research campus starting in November. The photo shows Luka Stärk receiving the award, with the chairs of the Scientific and Organizing Committees of the Competition, Giorgio Sartor and Bjørnar Luteberget, on his left and right, respectively. Congratulations to Carolin and Luka!

Luka Stärk and Carolin Scholl win the DISPLIB 2025 Competition for real-time train dispatching! #branch-and-bound algorithm for difficult #scheduling problems. #MODAL #MobilityLab @zuseinstitute.bsky.social
@mathplusberlin.bsky.social, @ Internat. Conf. Optim. & Decision Science, Milan 4 Sept 2025.

7 months ago 1 0 0 0
Detail - FG IBBB

16 Sept 2025, HTW Berlin: „Mathematical Optimization for the Energy Transition“ & „Introduction to Methods of Explainable AI" #XAI
Workshop Wirtschaftsmathematik meets Informatics at HTW with MODAL contributions by Janina Zittel and Ambros Gleixner
fg-ibbb.gi.de/veranstaltun...

7 months ago 0 0 0 0
Advertisement
Dr. Ina Czyborra, Senatorin für Wissenschaft, Gesundheit und Pflege des Landes Berlin, Prof. Dr. Günter Ziegler, Präsident der Freien Universität Berlin, Dr. Karsten Hess, Referatsleiter Universum und Materie im BMFTR, Bundesministerium für Forschung, Technologie und Raumfahrt at the opening of MODAL's third funding period, @freieuniversitaet.bsky.social @gmziegler.bsky.social‬ @zuseinstitute.bsky.social @mathplusberlin.bsky.social @senbjf.bsky.social‬ ( c ) ZIB

Dr. Ina Czyborra, Senatorin für Wissenschaft, Gesundheit und Pflege des Landes Berlin, Prof. Dr. Günter Ziegler, Präsident der Freien Universität Berlin, Dr. Karsten Hess, Referatsleiter Universum und Materie im BMFTR, Bundesministerium für Forschung, Technologie und Raumfahrt at the opening of MODAL's third funding period, @freieuniversitaet.bsky.social @gmziegler.bsky.social‬ @zuseinstitute.bsky.social @mathplusberlin.bsky.social @senbjf.bsky.social‬ ( c ) ZIB

MODAL - Smart Solutions for the Digital Society - BMFTR, German Federal Ministry of Research, Technology, and Space highlights MODAL's third funding period:
www.forschungscampus.bmbf.de/aktuelles/st...

8 months ago 1 0 0 0
Image: M.H. at META2025, Malaga, upcoming presentation: "Physics Informed Bayesian Optimization For Efficient Solution Of Inverse Problems In Nanophotonics" Abstract: We introduce physics-informed multi-channel Bayesian optimization for inverse design problems in nanophotonics and nanometrology. By learning vector-valued physical properties of numerically simulated systems, from which the final cost is constructed, our approach achieves faster convergence compared to other optimization methods, enhancing efficiency in solving inverse problems across engineering and physics.

Image: M.H. at META2025, Malaga, upcoming presentation: "Physics Informed Bayesian Optimization For Efficient Solution Of Inverse Problems In Nanophotonics" Abstract: We introduce physics-informed multi-channel Bayesian optimization for inverse design problems in nanophotonics and nanometrology. By learning vector-valued physical properties of numerically simulated systems, from which the final cost is constructed, our approach achieves faster convergence compared to other optimization methods, enhancing efficiency in solving inverse problems across engineering and physics.

Martin Hammerschmidt, CEO @JCMwave.bsky.social, presents MODAL-#NanoLab results at the 15th International Conference on Metamaterials, Photonic Crystals and Plasmonics, META 2025, Malaga, Spain, 23 July 2025. #MachineLearning #BayesianOptimization @cnoatzib.bsky.social‬,
Inv Talk in Session 2A10.

8 months ago 3 0 0 0
Enrico Bortoletto introduced two novel geometric perspectives on the periodic event scheduling problem, that was hitherto studied in terms of periodic tensions. In the periodic timetabling space, the set of feasible solutions turns out to be a collection of polytropes, i.e., tropical polytopes that are also polytopes in the Euclidean sense. These polytropes have a neighborhood structure that arises from flips in periodic tensions, and that in turn gives rise to a new tropical neighborhood search algorithm. This method effectively complements existing methods and fueled the discovery of new best solutions to five of the twenty-two instances of the PESPLIB, a collection notoriously difficult benchmark testset for periodic timetabling. In the cycle offset space, the fractional solutions produce a cographic zonotope feature a fine tiling, in which the maximal tiles correspond in a duality relation with the feasible timetables. Furthermore, the volume of the cographic zonotope equals the number of spanning trees, and a scaled version provides a lower bound for the width of a cycle basis, two result that are interesting in their own right. Cycles are also the key to deal with constraints on the occupation of infrastructure elements that ensure, e.g., that at any time only one train is scheduled to be at any platform; these inability to deal with such constraints has been a major problem in practical applications, in particular, in the timetabling for construction sites. By a canonical extension of the standard periodic timetabling formulation, such constraints can now be dealt with efficiently, as the thesis demonstrates at the example of scenarios at the S-Bahn Berlin. Photo (l to r) P. Schiewe, C Haase, E Bortoletto, R Borndörfer, A Freyer. https://www.zib.de/research/projects/modal-mobilitylab https://www.zib.de/research/projects/algebraic-and-tropical-methods-periodic-timetabling https://mathplus.de/research-2/application-areas/aa3-next-generation-networks/aa3-8/

Enrico Bortoletto introduced two novel geometric perspectives on the periodic event scheduling problem, that was hitherto studied in terms of periodic tensions. In the periodic timetabling space, the set of feasible solutions turns out to be a collection of polytropes, i.e., tropical polytopes that are also polytopes in the Euclidean sense. These polytropes have a neighborhood structure that arises from flips in periodic tensions, and that in turn gives rise to a new tropical neighborhood search algorithm. This method effectively complements existing methods and fueled the discovery of new best solutions to five of the twenty-two instances of the PESPLIB, a collection notoriously difficult benchmark testset for periodic timetabling. In the cycle offset space, the fractional solutions produce a cographic zonotope feature a fine tiling, in which the maximal tiles correspond in a duality relation with the feasible timetables. Furthermore, the volume of the cographic zonotope equals the number of spanning trees, and a scaled version provides a lower bound for the width of a cycle basis, two result that are interesting in their own right. Cycles are also the key to deal with constraints on the occupation of infrastructure elements that ensure, e.g., that at any time only one train is scheduled to be at any platform; these inability to deal with such constraints has been a major problem in practical applications, in particular, in the timetabling for construction sites. By a canonical extension of the standard periodic timetabling formulation, such constraints can now be dealt with efficiently, as the thesis demonstrates at the example of scenarios at the S-Bahn Berlin. Photo (l to r) P. Schiewe, C Haase, E Bortoletto, R Borndörfer, A Freyer. https://www.zib.de/research/projects/modal-mobilitylab https://www.zib.de/research/projects/algebraic-and-tropical-methods-periodic-timetabling https://mathplus.de/research-2/application-areas/aa3-next-generation-networks/aa3-8/

Enrico Bortoletto successfully defended his PhD dissertation “Geometric Advances and Infrastructure Awareness in Periodic Timetabling” at the Institute of Mathematics at Freie Universität Berlin on July 14, 2025, @zuseinstitute.bsky.social @mathplusberlin.bsky.social @fc-modal.bsky.social‬

9 months ago 2 1 0 0
Image: Solar TAP Online Workshop on Digital Twins for Emerging Photovoltaics, https://solartap.de/events/solar-tap-online-workshop-on-digital-twins-for-emerging-photovoltaics/

Image: Solar TAP Online Workshop on Digital Twins for Emerging Photovoltaics, https://solartap.de/events/solar-tap-online-workshop-on-digital-twins-for-emerging-photovoltaics/

Solar TAP Online Workshop on Digital Twins for Emerging #Photovoltaics, 30.06.-01.07.2025: solartap.de/events/ features a session on Optical Simulations with #JCMsuite by #JCMwave #SolarTAP @fc-modal.bsky.social‬ @kit.edu‬ @crockstuhl.bsky.social

9 months ago 2 1 0 0
Post image

On the occasion, Thorsten Koch crated cocktails in the colors of Forschungscampus MODAL

10 months ago 3 0 0 0

Opening speeches given by Ina Czyborra, Karsten Hess, @gmziegler.bsky.social‬, Carsten Busch, Thomas Herrmann, Thomas Meißner, Ralf Borndörfer, and Sebastian Pokutta! Highlights from the labs: EnergyLab, MedLab, MobilityLab, NanoLab, HPCLab, SynLab

10 months ago 2 1 1 0
Post image

Opening of MODAL / Phase3 on 16 June 2025 & Inauguration party of new #FUHUB labs and offices at #FUBIC @zuseinstitute.bsky.social @freieuniversitaet.bsky.social @mathplusberlin.bsky.social @htwberlin.bsky.social‬

10 months ago 4 0 1 1
Gioni Mexi, Deborah Hendrych, Sébastien Designolle, Mathieu Besançon, and Sebastian Pokutta were awarded the Land-Doig Competition Prize 2025 at the Mixed-Integer Programming Workshop 2025 for the best contribution to the heuristic solution of Mixed-Integer Quadratic Optimization Problems (MIQP and MIQCQP). The competition focused on the development of innovative primal heuristics that can quickly generate high-quality feasible solutions. The awarded contribution extended Boscia.jl - developed at the Zuse Institute Berlin - with novel Frank-Wolfe-based large neighborhood search heuristics, as well as gradient- and decomposition-based heuristics, specifically designed for quadratic optimization problems. The method found feasible solutions for 88 out of 95 benchmark instances, on average in just 12.8 seconds, and achieved a mean optimality gap of 5.1%. In addition, it improved eight best-known solutions in the QPLIB benchmark library. The photo shows, from left to right: Jan Kronqvist, Assistant Professor at KTH Royal Institute of Technology in Stockholm and member of the competition's organizing team, and Gioni Mexi.

Gioni Mexi, Deborah Hendrych, Sébastien Designolle, Mathieu Besançon, and Sebastian Pokutta were awarded the Land-Doig Competition Prize 2025 at the Mixed-Integer Programming Workshop 2025 for the best contribution to the heuristic solution of Mixed-Integer Quadratic Optimization Problems (MIQP and MIQCQP). The competition focused on the development of innovative primal heuristics that can quickly generate high-quality feasible solutions. The awarded contribution extended Boscia.jl - developed at the Zuse Institute Berlin - with novel Frank-Wolfe-based large neighborhood search heuristics, as well as gradient- and decomposition-based heuristics, specifically designed for quadratic optimization problems. The method found feasible solutions for 88 out of 95 benchmark instances, on average in just 12.8 seconds, and achieved a mean optimality gap of 5.1%. In addition, it improved eight best-known solutions in the QPLIB benchmark library. The photo shows, from left to right: Jan Kronqvist, Assistant Professor at KTH Royal Institute of Technology in Stockholm and member of the competition's organizing team, and Gioni Mexi.

ZIB team of Gioni Mexi and Coauthors wins the Land-Doig MIP Competition at #mip2025 @fc-modal.bsky.social‬ @mathplusberlin.bsky.social
www.zib.de/news/team-wi...

10 months ago 13 2 1 0
Advertisement

@spie.org‬ Digital Optical Technologies 2025 #LaserMunich: Visit Talk by Ivan Sekulic on Physics informed Bayesian optimization for inverse design of diffractive optical elements! @jcmwave.bsky.social @zuseinstitute.bsky.social @fc-modal.bsky.social‬ #NumericalOptimization #Photonics

10 months ago 3 2 1 0
Post image

Google DeepMind’s #AlphaEvolve recently set records on 13 notoriously difficult math problems. At MODAL we were curious: Could we tackle those with mathematical optimization instead of AI heuristics? Read the FICO blog to learn how we improved on DeepMind’s results:
www.fico.com/blogs/best-g...

10 months ago 0 0 0 0
Post image
10 months ago 0 0 0 0
Post image Post image

MODALs new industrial partner Stryker Berlin GmbH visited MODAL at ZIB together with their Freiburg colleagues from Stryker Leibinger GmbH & Co KG, Patient Specific Solutions. Research in #Medicine #LifeSciences #3Dvisualizations.

10 months ago 0 0 0 0

The opening of Phase 3 of the MODAL research campus will take place on June 16, 2025 at the FUHUB in Berlin.

10 months ago 0 0 0 0

The research campus is hosted by ZIB, located at the campus of the FU in Berlin, and conducts research and development of digital systems for the optimization of data-driven processes in the fields of energy, health, mobility and communication.

10 months ago 1 0 0 0

MODAL is a partnership project of research partners @zuseinstitute.bsky.social & @freieuniversitaet.bsky.social‬ with >30 industrial partners, funded by the Federal Ministry of Research, Technology and Space within the “Research Campus – Public-Private Partnership for Innovation” funding initiative.

10 months ago 2 0 1 0