Advanced Algorithms and Optimization Models Supported by Mathematical Theory (OptimaAI)

Basic Information

The OptimaAI project focuses on the development of innovative artificial intelligence algorithms and optimization models grounded in modern mathematical theories. The aim of the project is to build energy-efficient and scalable solutions for the analysis of complex data in bioinformatics, natural language processing, visual tracking, and industrial optimization. A key component of the project is the strong synergy between applied AI technologies (deep learning, NLP, edge AI) and fundamental mathematical disciplines (combinatorics and discrete mathematics, number theory, and group theory).

The project is structured around four work packages: WP1 – Development of advanced bioinformatics tools for single-cell transcriptome analysis; WP2 – Development of AI models for large-scale datasets and intelligent real-time camera signal processing; WP3 – Deep learning for combinatorial optimization and graph theory; and WP4 – Solving mathematical problems applicable to optimization through heuristics and metaheuristics using theoretical, combinatorial, and discrete approaches.

Expected results include open-source software, high-impact scientific publications, and the transfer of results into applications relevant to healthcare, industry, and the digital transition.

Project duration: 1.10.2025 – 30.9.2029.

Publications

1. M. Jukić Bokun, I. Soldo, Triangular D(−1)-tuples,, Bull. Math. Soc. Sci. Math. Roumanie, to appear.
2. K. Sabo, R. Scitovski, D. Grahovac, Š. Ungar, Mahalanobis clustering for color image segmentation and application to skin lesion classification,  subbmited.
3. D. Krupić, D. Matijević, N. Šuvak, J. Maltar, D. Ševerdija, Evaluating the agreement between human preferences, GPT-4V and Gemini Pro Vision assessments: Can AI recognize what people might like?, Computers in Human Behavior: Artificial Humans 6 (2025)

Conferences

1. J. Maltar, I. Marković, D. Matijević, I. Petrović: Deep Recurrent Visual Place Recognition Based on Softmax Fine-Tuned Image Representations, MOTSP 2026, Zadar, Croatia
2. T. Prusina, J. Benić, D. Ševerdija, D. Matijević: MULTI-PERSON 2D HUMAN POSE ESTIMATION: A BENCHMARK FOR REAL-TIME APPLICATIONS, MOTSP 2026, Zadara, Croatia
   git: https://github.com/JurajBenic/HPEBench2D
3. T. Prusina, J. Benić: Edge-Based Bearing Condition Monitoring Using Jetson AGX Orin and Neural Networks, MOTSP 2026, Zadara, Croatia
4. L. Borozan, B. Borozan, D. Ševerdija: Efficient Checkpointing via Object Serialization in C# Applications, 2026 MIPRO 49th ICT and Electronics Convention, Opatija, Croatia, 2026.
   git: https://github.com/LukaBorozan/CheckpointingSystem
5. J. Sabljo, M. Đumić, M. Đurasević: Analysis of Training Models for Evolving Heuristics in the Unrelated Parallel Machines Scheduling Problem with Additional Resources, 2026 MIPRO 49th ICT and Electronics Convention, Opatija, Croatia, 2026.
   git: https://github.com/jsabljo/UPMSP-PR.git
6. Josipa Sabljo, Mateja Ðumić, and Marko Ðurasević. 2026. Evolving Dispatching Rules for the Unrelated Parallel Machines Scheduling Problem with Precedence and Resource Constraints with Genetic Programming. In Genetic and Evolutionary Computation Conference (GECCO ’26), July 13–17, 2026, San Jose, Costa Rica. ACM, New York, NY, USA, 9 pages.
   https://doi.org/10.1145/3795095.3805192

Mobility

• Bartol Borozan, Faculty of Informatics and Data Science, University of Regensburg, Germany, 1.10.2025. – 30.06.2026.
• Luka Borozan, Faculty of Informatics and Data Science, University of Regensburg, Germany, 24.3.2026.-28.3.2026.

News