Computational Intelligence | Department of Intelligent Systems

Computational Intelligence

Computational intelligence is a study of stochastic search, optimization and learning methods, inspired by physical and biological systems. Research in this area at the Department of Intelligent Systems focuses on the evolutionary computation methods. We study extensions of evolutionary algorithms for multiobjective optimization and their parallelization, and apply these algorithms in engineering design and optimization problems.


  • SYNERGY – Synergy for Smart Multi-Objective Optimization
    Demanding real-world optimization problems with multiple objectives can be solved more efficiently if parallelization is used to execute a number of simulations simultaneously and if the the time-consuming simulations are partly replaced by accurate surrogate models. In this twinning project, JSI’s Computer Systems and Intelligent Systems departments aim to strengthen their research and innovation potential in parallelization and surrogate modelling and to explore the potential of combining the two techniques in smart multi-objective optimization in collaboration with the University of Lille and Cologne University of Applied Sciences.
  • Advanced methodology of evolutionary multi- and many-objective optimization for real-world applications
    This collaboration between Japanese and Slovenian researchers aims at advancing the methodology of evolutionary multi- and many-objective optimization for highly relevant real-world applications. It considers three important topics: (i) state-of-the-art evolutionary algorithms design and development, (ii) surrogate models and visualization tools, and (iii) real-world applications. Unlike most other research efforts, our research aims to unify these approaches to effectively tackle the problems observed in complex application domains that are multi- and many-objective, computationally expensive, require surrogate models, and benefit from domain knowledge. Furthermore, we plan to incorporate visualization of high-dimensional spaces to help in the analysis of optimization and guide the evolutionary search process.
  • Incorporating real-world problems into the benchmarking of multiobjective optimizers
    Only a few real-world multiobjective optimization problems are freely available for research purposes. Therefore, there is an urgent need to collect real-world problems, models of real-world problems and more realistic synthetic benchmark problems into an open benchmark suite that could be used by any researcher in multiobjective optimization. The idea of this project is to extend the state-of-the-art open-source COCO platform by incorporating real-world problems and their properties in order to bridge the gap between research and application in multiobjective optimization.


  • Visualization in Multiobjective Optimization
    Tutorial at GECCO 2016 and CEC 2017 by Bogdan Filipič and Tea Tušar
    This tutorial provides a comprehensive overview of methods used in multiobjective optimization for visualizing either individual approximation sets or the probabilistic distribution of multiple approximation sets through the empirical attainment function.
    download handouts from GECCO 2016
  • Deploying Evolutionary Computation in Industry: Challenges and Lessons Learned
    Tutorial at CEC 2015 by Bogdan Filipič
    This tutorial presents the challenges faced and lessons learned in designing evolutionary algorithms for industrial optimization problems that go beyond the textbook knowledge on optimization and evolutionary computation. It starts with defining the scope of the presentation and providing motivating examples of industrial applications with diverse characteristics. The core of the tutorial is a systematic analysis of the potential challenges illustrated with practical situations, followed by an overview of the lessons learned in dealing with these challenges when deploying evolutionary algorithms. It is the ambition of the tutorial to both promote evolutionary computation as a practical problem-solving methodology and contribute to bridging the gap between the algorithm designers and end-users.

Past Projects

Completed Doctoral Research Projects

  • Surrogate Based Multiobjective Algorithm for Solving Computationally Expensive Numerical Problems Based on Relations Under Uncertainty
    Miha Mlakar, supervisor: Bogdan Filipič
    The goal of this doctoral research is to develop effective and efficient multiobjective optimization algorithm for solving hard numerical problems, where solution evaluation is computationally expensive. To achieve this goal the algorithm builds a surrogate model of the original objective function. With this surrogate model some solutions are then approximated and are used in the optimization process.
    Project accomplished with a successful Ph.D. dissertation defense in April 2015.
  • Visualizing Solution Sets in Multiobjective Optimization
    Tea Tušar, supervisor: Bogdan Filipič
    This doctoral research addresses two distinct tasks in visualization in multiobjective optimization—visualization of approximation sets and visualization of empirical attainment functions (EAFs). In the first task it aims at developing a method for visualizing approximation sets that preserves the Pareto dominance relation between as many visualized points as possible. In addition, it presents a comprehensive review of the existing visualization methods used in evolutionary multiobjective optimization, showing their outcomes on two novel 4D benchmark approximation sets. In the second task the research addresses the visualization of exact as well as approximated 3D EAF values and differences in these values provided by two competing multiobjective optimization algorithms.
    Project accomplished with a successful Ph.D. dissertation defense in September 2014.
    view site
  • Multiobjective Discovery of Driving Strategies
    Erik Dovgan, supervisor: Bogdan Filipič, co-supervisor: Matjaž Gams
    The goal of this doctoral research is to improve the methodology for discovering driving strategies. We proposed and developed a multiobjective optimization algorithm that searches for driving strategies using a black-box vehicle model and minimizes the traveling time and the fuel consumption. Tests on data from real-world routes show that the proposed algorithm finds better driving strategies than the existing algorithms.
    Project accomplished with a successful Ph.D. dissertation defense in January 2014.
    view site

13th International Conference on Parallel Problem Solving from Nature

PPSN 2014The 13th International Conference on Parallel Problem Solving from Nature, PPSN 2014, was co-organized by the Department of Intelligent Systems and held at the Ljubljana Exhibition and Convention Centre on September 13-17, 2014. The conference  brought together researchers and practitioners in the field of natural computing, i.e., the study of computational systems that use ideas and get inspiration from natural systems, including biological, ecological, physical, chemical, and social systems. The conference proceedings are published in Lecture Notes in Computer Science by Springer.