Research
Research Topics
Below is a summary of my main research interests, highlighting representative publications. For a comprehensive list of my publications, please refer to the publications page.
Constrained Multiobjective Optimization
My research explores how constraints impact multiobjective optimization problems. Specifically, I focus on:
- Characterizing constrained problems through their feature and performance spaces
- Visualizing how constraints reshape the problem landscape
- Predicting algorithm performance on constrained multiobjective problems
Understanding Multiobjective Optimization Problems
I am interested in improving the understanding of multiobjective optimization problems, particularly by:
- Analyzing local correlations between objectives
- Investigating the latent structure of benchmark problems
Benchmarking Methodology and Problems
In collaboration with colleagues from the RandOpt Team at Inria, France, we actively advance benchmarking practices and the development of new benchmark optimization problems. This includes:
- The COCO platform for comparing optimization algorithms
- The anytime performance assessment methodology, including when algorithms are budget-dependent
- Specific suites of artificial benchmark problems, such as bbob-biobj, bbob-mixint, and bbob-largescale
In addition, through other international collaborations, we have introduced suites of game-based benchmark problems (see also this dedicated website) and explored a methodology for creating a diverse problem suite.
Visualization Techniques
Visualization techniques play a central role in my research, offering clear insights into the properties of optimization problems and the internal dynamics of optimization algorithms. I have worked on:
- Visualizing high-dimensional solution sets with prosections
- Visualizing algorithm performance through Empirical Attainment Functions (EAFs)
- Classifying visualization methods through a structured taxonomy
- Providing visual interpretations of algorithm behavior using Diversity and Usage maps (DU maps) and population evolution visualizations
- Visualizing the 2-D problem landscapes of the bbob-biobj problems with various methods
Real-World Applications
Over the years, I have tackled a variety of real-world optimization problems:
Scheduling for Renewable Energy Integration
- Balancing electricity supply and demand
- Optimizing demand response strategies
Engineering and Manufacturing
- Exploring electric motor design optimization through both a single-step method comparison and a multi-step process tailored to achieving robust designs
- Improving the steel casting process
- Tunnel design and horizontal alignment optimization
- Enhancing commutator manufacturing
Transport and Logistics
- Optimizing commodity transportation
- Analyzing efficient driving strategies
- Scheduling fieldwork
Other Domains
- Optimizing public health strategies during the COVID-19 pandemic
- Optimizing procedural content generation for games
- Space exploration via trajectory design optimization
Additionally, we have identified properties of real-world optimization problems with a survey (see also this dedicated website for the results).
Multiobjective Optimization Algorithms
I have contributed to developing and analyzing multiobjective optimization algorithms based on differential evolution, notably:
- The DEMO (Differential Evolution for Multiobjective Optimization) algorithm
- GP-DEMO (Gaussian-Process-based DEMO) for tackling problems with computationally expensive solution evaluations
- BBDEMO (Black-Box DEMO), inspired by the BBDE and COMO-CMA-ES algorithms, made for solving mixed-integer problems
Projects
To see a list of current and past projects, please see the Computational Intelligence group website.
Software
I have contributed to the development of the following software tools and packages:
The COCO Platform
COCO (Comparing Continuous Optimizers) is a software platform for a systematic and sound comparison of continuous and mixed optimization algorithms.
The moarchiving Library
The moarchiving Python library implements a multi-objective non-dominated archive for 2, 3 or 4 objectives, providing easy and fast access to multiple hypervolume indicators.
The OPI Tool
OPI (Optimization Problem Inspector) is a tool for analysis of industrial optimization problems and their solutions.
Talks
Ideals and realities of benchmarking in evolutionary multiobjective optimization
Keynote at EvoStar 2025, 23 April 2025, Trieste, ItalyData Visualization
Three invited talks for an internal workshop of the A training program for improving research on illiberal systems and finding ways to build more robust democracies – SOS4Democracy project, 18, 20 and 25 March 2025, onlineTwo suites of mixed-integer optimization problems
Invited talk at the Good Benchmarking Practices for Evolutionary Computation Workshop, GECCO 2023, 16 July 2023, Lisbon, PortugalSome issues in benchmarking multiobjective optimization algorithms
Invited talk at Dagstuhl Seminar 23251, Challenges in Benchmarking Optimization Heuristics, 19 June 2023, Schloss Dagstuhl, WadernA real-world perspective on benchmarking optimization algorithms
Invited talk at the Good Benchmarking Practices for Evolutionary Computation Workshop, GECCO 2022, 9 July 2022, Boston, USARoadblocks to finding optimal tunnel alignments with evolutionary algorithms Invited talk at Evolutionary Computation in Practice, GECCO 2021, 12 July 2021, online
Optimiranje tras predorov z evolucijskimi algoritmi
Invited talk at the Faculty of Electrical Engineering and Computer Science, University of Maribor, 10 December 2020, onlineVisualization in evolutionary computation
Invited talk at the Workshop on Evolutionary and Population-based Optimization, WEPO 2020, 26 November 2020, onlineTowards using real-world problems for benchmarking multiobjective optimizers
Invited talk at Cologne University of Applied Sciences, 25 May 2018, Gummersbach, GermanyVečkriterijska optimizacija z evolucijskimi algoritmi
Two invited talks at the Faculty of Mathematics and Physics, University of Ljubljana, 31 May and 2 June 2017, Ljubljana, SloveniaAn introduction to data visualization
Invited talk at the Joint ICTP-IAEA Workshop on Environmental Mapping, 13 March 2017, Trieste, ItalyAn overview of visualization methods for multiobjective optimization
Invited talk at the Alliance Manchester Business School, University of Manchester, 22 September 2016, Manchester, UK
Tutorials
Benchmarking Multobjective Optimizers 2.0
Tutorial with Dimo Brockoff at GECCO 2021, GECCO 2022, EMO 2023 and GECCO 2023Bibtex
@inproceedings{Brockhoff21TutorialGECCO, author = {Dimo Brockhoff and Tea Tu{\v s}ar}, title = {Benchmarking multiobjective optimizers 2.0}, booktitle = {{GECCO} '21: Genetic and Evolutionary Computation Conference, Companion Volume}, pages = {664--668}, publisher = {{ACM}}, year = {2021}, doi = {10.1145/3449726.3461421} }@inproceedings{Brockhoff22TutorialGECCO, author = {Dimo Brockhoff and Tea Tu{\v s}ar}, title = {{GECCO} 2022 Tutorial on benchmarking multiobjective optimizers 2.0}, booktitle = {{GECCO} '22: Genetic and Evolutionary Computation Conference, Companion Volume}, pages = {1269--1309}, publisher = {{ACM}}, year = {2022}, doi = {10.1145/3520304.3533635} }@inproceedings{Brockhoff23TutorialGECCO, author = {Dimo Brockhoff and Tea Tu{\v s}ar}, title = {{GECCO} 2023 Tutorial on Benchmarking Multiobjective Optimizers 2.0}, booktitle = {Companion Proceedings of the Conference on Genetic and Evolutionary Computation, {GECCO} 2023}, pages = {1183--1212}, publisher = {{ACM}}, year = {2023}, doi = {10.1145/3583133.3595060} }
Visualization in Multiobjective Optimization
Tutorial with Bogdan Filipič at GECCO 2016, CEC 2017, GECCO 2018, GECCO 2019 and GECCO 2020
See the supplementary material page for additional resourcesBibtex
@inproceedings{Filipic16Tutorial, author = {Bogdan Filipi{\v c} and Tea Tu{\v s}ar}, title = {Visualization in Multiobjective Optimization}, booktitle = {Companion Material Proceedings of the Genetic and Evolutionary Computation Conference, {GECCO} 2016}, pages = {735--751}, publisher = {{ACM}}, year = {2016}, doi = {10.1145/2908961.2926994} }@inproceedings{Filipic18Tutorial, author = {Bogdan Filipi{\v c} and Tea Tu{\v s}ar}, title = {Visualization in Multiobjective Optimization}, booktitle = {Companion Proceedings of the Genetic and Evolutionary Computation Conference, {GECCO} 2018}, pages = {858--879}, publisher = {{ACM}}, year = {2018}, doi = {10.1145/3205651.3207891} }@inproceedings{Filipic19Tutorial, author = {Bogdan Filipi{\v c} and Tea Tu{\v s}ar}, title = {Visualization in Multiobjective Optimization}, booktitle = {Companion Proceedings of the Genetic and Evolutionary Computation Conference, {GECCO} 2019}, pages = {951--974}, publisher = {{ACM}}, year = {2019}, doi = {10.1145/3319619.3323374} }@inproceedings{Filipic20Tutorial, author = {Bogdan Filipi{\v c} and Tea Tu{\v s}ar}, title = {Visualization in Multiobjective Optimization}, booktitle = {{GECCO} '20: Genetic and Evolutionary Computation Conference, Companion Volume}, pages = {775--800}, publisher = {{ACM}}, year = {2020}, doi = {10.1145/3377929.3389867} }
Awards
Taras award for successful collaboration of the economy and the research and development environment in the field of innovation, development and technology, Industrial Forum IRT 2023, Portorož, Slovenia
Members of the Department of Intelligent Systems (Tea Tušar, Aljoša Vodopija, Jordan Cork, Bogdan Filipič), Department of Computer Systems (Peter Korošec) and the company Mahle Electric Drives Slovenija, d.o.o.: Development of an electric motor for the automotive industry with an innovative simulation-based optimization procedureBest poster award at EvoStar 2023, Brno, Czechia
Andrejaana Andova, Tobias Benecke, Harald Ludwig, Tea Tušar: Towards constructing a suite of multi-objective optimization problems with diverse landscapesBest paper award at the Slovenian Conference on Artificial Intelligence, 26th International Multiconference Information Society, IS 2023, Ljubljana, Slovenia
Tea Tušar, Peter Korošec, Bogdan Filipič: A multi-step evaluation process in electric motor designSecond place at the XPRIZE Pandemic Response Challenge, 2021
Team JSI vs. COVIDBest paper award at the XV International Conference ETAI 2021, Skopje, Northern Macedonia
Nina Reščič, Vito Janko, David Susič, Carlo De Masi, Aljoša Vodopija, Matej Marinko, Tea Tušar, Erik Dovgan, Matej Cigale, Anton Gradišek, Matjaž Gams, Mitja Luštrek: Finding efficient intervention plans against COVID-19Best paper award at Human-Computer Interaction in Information Society, 23rd International Multiconference Information Society, IS 2020, Ljubljana, Slovenia
Tea Tušar: Interactive visualization of the Slovenian budget with the Sankey diagram