
Kārlis Freivalds
Dr. sc. comp.
Senior Researcher
Dr. Kārlis Freivalds is a senior researcher at the Institute of Electronics and Computer Science. He has graduated from the University of Latvia (Riga, Latvia) where he obtained a doctoral degree in computer science. Kārlis is the author of 30+ scientific publications and 7 patent applications. He has been involved in various scientific and industrial projects in the areas of computer vision, artificial intelligence, graph algorithms, bioinformatics and quantum computing, where he has led research work as well as software development and deployment in production. The main areas of research are artificial intelligence, computer vision, robotics and machine learning, with a recent focus on deep neural networks, where he has developed several new neural network architectures for fast sequence processing, algorithmic tasks and NP-hard problem solving. He is also an associate professor at the Computer Science department of the University of Latvia, where he teaches programming of image processing and computer vision algorithms.
Related publications:
- Teikmanis, O., Leja, L., & Freivalds, K. Applying a Differentiable Physics Simulation to Move Objects with Fluid Streams. (2023). International Workshop on Embedded Digital Intelligence (IWoEDI’2023).
- E. Ozolins, K. Freivalds, A. Draguns, E. Gaile, R. Zakovskis, S. Kozlovics "Goal-aware neural SAT solver" 2022 International Joint Conference on Neural Networks (IJCNN), 1-8, IEEE.
- Gatis Melkus, Karlis Cerans, Karlis Freivalds, Lelde Lace, Darta Zajakina, Juris Viksna "Analysis of Dynamics and Stability of Hybrid System Models of Gene Regulatory Networks." The 12th International Conference on Computational Systems-Biology and Bioinformatics. 2021.
- Draguns, A., Ozoliņš, E., Šostaks, A., Apinis, M., & Freivalds, K. (2021). Residual Shuffle-Exchange Networks for Fast Processing of Long Sequences. Proceedings of the AAAI Conference on Artificial Intelligence, 35(8), 7245-7253.
- Ozolinš, K. Freivalds and A. Šostaks, "Matrix Shuffle- Exchange Networks for Hard 2D Tasks," 2021 International Joint Conference on Neural Networks (IJCNN), 2021, pp. 1-8, doi:10.1109/IJCNN52387.2021.9533919.
- Barzine, M.P., Freivalds, K., Wright, J.C., Opmanis, M., Rituma, D., Ghavidel, F.Z., Jarnuczak, A.F., Celms, E., Čerāns, K., Jonassen, I. and Lace, L., 2020. Using Deep Learning to Extrapolate Protein Expression Measurements. Proteomics, 20(21-22), p.2000009.
- Kārlis Freivalds, Emīls Ozoliņš, and Agris Šostaks. "Neural Shuffle-Exchange Networks-Sequence Processing in O (n log n) Time." In Advances in Neural Information Processing Systems, pp. 6626-6637. 2019.
- Karlis Martins Briedis, Karlis Freivalds. On-line Television Stream Classification by Genre. Baltic Jornal of.Modern Computing, Vol. 6 (2018), No. 3, pp. 235–246.
- Karlis Freivalds, Renars Liepins. "Improving the Neural GPU Architecture for Algorithm Learning." The ICML workshop Neural Abstract Machines & Program Induction v2 (NAMPI 2018).
- Arturs Sprogis, Karlis Freivalds, and Elita Cirule. "Implementing a Face Recognition System for Media Companies." In International Baltic Conference on Databases and Information Systems, pp. 328-341. Springer, Cham, 2018.
- Jans Glagoļevs, Kārlis Freivalds “Logo Detection in Images Using HOG and SIFT”, 5th IEEE Workshop on Advances in Information, Electronic and Electrical Engineering (AIEEE), Riga, Latvia, 2017, pp. 1-5.
- Kārlis Freivalds, Jans Glagoļevs “A Statistical Method for Object Counting”. Proceedings of 2017 International Conference on Digital Signal Processing (ICDSP 2017) Kuala Lumpur, Malaysia
- Jevgenijs Vihrovs, Krišjānis Prūsis, Kārlis Freivalds, Pēteris Ručevskis and Valdis Krebs "A Potential Field Function for Overlapping Point Set and Graph Cluster Visualization." International Joint Conference on Computer Vision, Imaging and Computer Graphics (pp. 136-152). Springer, 2015.
- Kārlis Freivalds, Jans Glagoļevs. "Graph Compact Orthogonal Layout Algorithm." 3rd International Symposium on Combinatorial Optimization, ISCO 2014, Springer LNCS Volume 8596, pp 255-266.
Related patents:
Projects
5 results
- Artificial Intelligence in Manufacturing leading to Sustainability and Industry 5.0 (AIMS5.0) 01.05.2023 – 30.04.2026 Completed
- Metodes, modeļi un rīki enerģētikas sistēmu viedai attīstībai (MESVA) 01.01.2025 – 31.03.2026 Completed
- Digitalization of Power Electronic Applications within Key Technology Value Chains (PowerizeD) 01.01.2023 – 31.12.2025 Completed
- Smart Materials, Photonics, Technologies, and Engineering Ecosystem (MOTE) 01.01.2023 – 31.12.2024 Completed
- Digital Technologies, Advanced Robotics and increased Cyber-security for Agile Production in Future European Manufacturing Ecosystems (TRINITY) 01.01.2019 – 31.12.2022 Completed
Publications
- 2026
Differentially Private Federated Learning for Remaining Useful Life Prediction(opens in a new window)
Arturs Nikulins, Kārlis Freivalds, Ivars Namatēvs, Kaspars Sudars, Audris Arzovs, Wilhelm Söderkvist Vermelin, Madhav Mishra, Kaspars Ozols.
Applied Sciences, Vol. 16 (6), p. 2784 (raksta numurs), MDPI
Abstract
Accurate remaining useful life (RUL) prediction is essential for the safe and cost-effective operation of safety-critical systems such as electronic components and engines. While data-driven machine learning approaches have demonstrated strong performance for RUL estimation, their effectiveness is limited by the lack of full run-to-failure data and by strict privacy and intellectual property constraints in industrial settings. Federated learning (FL) enables collaborative model training across multiple data owners without direct data sharing, but it does not, by itself, provide formal privacy guarantees and remains vulnerable to information leakage. This paper presents a privacy-preserving DP-enhanced FL setup for RUL prediction that combines federated learning with differential privacy (DP). We describe an end-to-end implementation based on the Opacus DP library, highlight practical challenges arising from the integration of DP into recurrent neural network architectures, and propose solutions to address them. Using two representative RUL datasets (CMAPSS and SiC MOSFET), we analyze the effect of DP noise on prediction performance and on the functional dependence between the predicted RUL and the already lived life feature. The results demonstrate that differential privacy can be integrated into federated RUL prediction with limited degradation in predictive performance, providing practical insights for deploying privacy-aware collaborative models in industrial environments.
Indexed in Scopus / WoS · Scopus percentile: 79 · SNIP: 0.956
- 2025
Applied AI for Adaptive Robotic Conveyor-Feeding (I)(opens in a new window)
Guntis V. Strazds, Katrīna Viltrake, Gergely Hollosi, Laura Leja, Arnis Priedītis, Diāna Dupļevska, Makss Meiers, Christos Chronis, Iraklis Varlamis, George Dimitrakopoulos, Kārlis Freivalds
1st IFAC Joint Conference on Computers, Cognition, and Communication, IFAC
Abstract
This paper presents a modular AI-driven robotic system for dynamic production line tasks, integrating real-time vision, adaptive motion planning, and making use of LLM-based code generation to facilitate reconfiguration for new products or tasks. The architecture of the overall system is described, as are design considerations and initial evaluation results for AI-based components for bottle pose estimation, tracking of moving sockets into which bottles need to be inserted, and training of a neural network for dynamic robot-motion control.
Indexed in Scopus / WoS · Scopus percentile: 36
- 2025
Differentiable Physics Training Method for Robot Motion Planning(opens in a new window)
Laura Leja , Guntis Vilnis Strazds, Christos Chronis, Iraklis Varlamis, Kārlis Freivalads
1st IFAC Joint Conference on Computers, Cognition, and Communication, IFAC
Details
Indexed in Scopus / WoS
- 2025
Controlling shape of flame with differentiable physics(opens in a new window)
Laura Leja, Kārlis Freivalds
Engineering for Rural Development, Vol. 24, p. 666 - 672, Latvia University of Life Sciences and Technologies
Abstract
The study focuses on creating symbol-shaped flame designs by optimizing flame dynamics, using machine learning techniques integrated with advanced combustion modelling. This research addresses the limitations of traditional pyrotechnic systems, which lack precise control over flame shapes, by introducing a method that uses differentiable physics simulations for shaping flames. The method uses the optimization framework PhiFlow and includes radiative heat loss and chemical kinetics for improved accuracy. Machine learning is used not only to train and adapt the model but also to determine the physical parameters that correspond to the actual observed behaviour of the flames. However, it is also important to mathematically and analytically evaluate the results obtained to ensure their accuracy and relevance. Results demonstrate the successful formation of custom flame shapes in a 2D simulated environment, with promising alignment to real-world flame stage projector behaviour. This approach enables control over flame shapes and dynamics, expanding creative possibilities in stage design and interactive art while proving valuable for applications requiring detailed control of dynamic systems, such as pyrotechnics, performance art, and educational demonstrations.
Indexed in Scopus / WoS · SNIP: 0.303
- 2024
Learning to Move Objects With Fluid Streams in a Differentiable Simulation(opens in a new window)
Kārlis Freivalds, Laura Leja, Oskars Teikmanis
2024 7th Iberian Robotics Conference, ROBOT 2024, IEEE
Abstract
We introduce a method for contactless manipulation of objects in three-dimensional space using controlled fluid streams. To achieve this, we train a neural network controller in a differentiable simulation and evaluate it in a simulated environment consisting of an 8 × 8 grid of vertical emitters. By carrying out various horizontal displacement tasks such as moving objects to specific positions while reacting to external perturbations, we demonstrate that a controller, trained with a limited number of iterations, can generalise to longer episodes and learn the complex dynamics of fluid-solid interactions. Importantly, our approach requires only the observation of the manipulated object's state, paving the way for the development of physical systems that enable contactless manipulation of objects using air streams.
Indexed in Scopus / WoS
Patents
2015 European Patent No. EP2700054 B1
System and method for video-based vehicle detection
Inventors:K.Freivalds, R.Kadikis, M.Greitans
Description and link
View the patent on Espacenet(opens in a new window)EP2700054 (type: A1, date: 26.02.2014; type: B1, date: 22.04.2015) “System and method for video-based vehicle detection”. Inventors: K.Freivalds, R.Kadikis, M.Greitans. Applicant:
Institute of Electronics and Computer Sciences of Latvia.