Igor Škrjanc
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Igor Škrjanc has a total of 83 co-authors. Here are the top co-authors by number of shared publications:
- Gorazd Karer
University of Ljubljana
18 shared publications - Sašo Blažič
University of Ljubljana
12 shared publications - Daniel Leite
Department of Computer Science, University of Paderborn, 33098 Paderborn, Germany
10 shared publications - Gregor Klančar
Faculty of Electrical Engineering, University of Ljubljana, 1000 Ljubljana, Slovenia
8 shared publications - Edwin Lughofer
Institute for Mathematical Methods in Medicine and Data Based Modeling, Johannes Kepler University, Linz, Austria
7 shared publications - Miha Ožbot
Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia
7 shared publications
Latest Publications
- Open Access
- 13 October 2026
- Published by Springer Nature in Evolving Systems
- Vol. 17 (4) , 123
Evolving federated density peak clustering with shuffler-based anonymization
In this paper, we present a novel one-shot federated clustering approach based on evolving fuzzy clustering and shuffler-based anonymization. In federated clustering, each round reveals additional model information to the server and can increase the opportunity to infer properties of local datasets, even when raw samples are never shared. Consequently, fewer client–server exchanges and unlinkability are important, so transmitted information cannot be attributed to any single client. Data-stream clustering algorithms, which learn in a single pass over the data, align well with the one-shot federated learning paradigm, where a single communication round can be sufficient. The proposed Evolving Federated Density Peak Clustering (eFDPC) method combines evolving fuzzy data-stream clustering and density peak clustering in a federated setting. Each client independently evolves micro-clusters from private data mini-batches through cluster addition and merging. To limit communication, density peak clustering is applied to the learned micro-clusters and only the selected peaks are transmitted, rather than the full local models. A shuffler-based layer is introduced between the clients and the server to anonymize contributions prior to reconstruction and global peak selection. We consider a peak-only variant, eFDPC- $$\mathcal {P}$$ , as well as eFDPC- $$\mathcal {G}$$ with server-side label propagation. Experiments on synthetic and real-world benchmark datasets show that eFDPC outperforms recent federated clustering baselines in one-shot settings and scales to higher-dimensional, non-IID data.- 13 October 2026
- Published by Springer Nature in Evolving Systems
- Vol. 17 (4) , 123
Open AccessEvolving federated density peak clustering with shuffler-based anonymization
In this paper, we present a novel one-shot federated clustering approach based on evolving fuzzy clustering and shuffler-based anonymization. In federated clustering, each round reveals additional model information to the server and can increase the opportunity to infer properties of local datasets, even when raw samples are never shared. Consequently, fewer client–server exchanges and unlinkability are important, so transmitted information cannot be attributed to any single client. Data-stream clustering algorithms, which learn in a single pass over the data, align well with the one-shot federated learning paradigm, where a single communication round can be sufficient. The proposed Evolving Federated Density Peak Clustering (eFDPC) method combines evolving fuzzy data-stream clustering and density peak clustering in a federated setting. Each client independently evolves micro-clusters from private data mini-batches through cluster addition and merging. To limit communication, density peak clustering is applied to the learned micro-clusters and only the selected peaks are transmitted, rather than the full local models. A shuffler-based layer is introduced between the clients and the server to anonymize contributions prior to reconstruction and global peak selection. We consider a peak-only variant, eFDPC- $$\mathcal {P}$$ , as well as eFDPC- $$\mathcal {G}$$ with server-side label propagation. Experiments on synthetic and real-world benchmark datasets show that eFDPC outperforms recent federated clustering baselines in one-shot settings and scales to higher-dimensional, non-IID data. - Open Access
- 30 September 2026
- Published by Springer Nature in Journal of Control, Automation and Electrical Systems
- p. 1-18
Adaptive Fuzzy LMS Control for Positioning and Anti-Sway of an Overhead Crane System
This paper formulates an adaptive fuzzy control approach for the overhead crane positioning and anti-sway problem. The crane is treated as an underactuated nonlinear system with four states, one horizontal force input, and two controlled outputs: trolley position and load swing angle. The introductory part of the paper is extended with a crane control-oriented literature overview that distinguishes open-loop input shaping and command smoothing, model-based feedback control, fuzzy and LMI-based control, data-driven and model-free adaptive control, reinforcement learning, and machine learning-based approaches. The control design is then developed from a compact Takagi–Sugeno fuzzy state-space representation of the nonlinear cart–pendulum crane dynamics. The fuzzy model is obtained by sector nonlinearity and local approximation in fuzzy partition spaces, resulting in two local rules scheduled by the load swing angle. A parallel distributed fuzzy controller is parameterized as a linear-in-the-parameters control law, and its local parameters are adapted online by a teacher-based normalized least mean squares (LMS) rule. The adaptive controller learns the control signal of a virtual supervisory fuzzy controller and then remains active as the controller in the test phase. Projection, leakage regularization, force saturation, and a smooth transfer from the supervisory signal to the learned LMS signal are included to obtain an implementation-oriented closed-loop formulation. The proposed approach is intended as a physically interpretable bridge between robust fuzzy crane modeling and data-driven adaptive control. The novelty is clarified as the integration of a compact fuzzy crane model, a supervisory teacher signal, normalized LMS learning, leakage/projection, input saturation, and command-governed anti-sway tuning in one simulation-oriented framework. The study is explicitly presented as a numerical benchmark; formal nonlinear stability certification and laboratory validation are identified as future work rather than as claims of the present contribution.
- 30 September 2026
- Published by Springer Nature in Journal of Control, Automation and Electrical Systems
- p. 1-18
Open AccessAdaptive Fuzzy LMS Control for Positioning and Anti-Sway of an Overhead Crane System
This paper formulates an adaptive fuzzy control approach for the overhead crane positioning and anti-sway problem. The crane is treated as an underactuated nonlinear system with four states, one horizontal force input, and two controlled outputs: trolley position and load swing angle. The introductory part of the paper is extended with a crane control-oriented literature overview that distinguishes open-loop input shaping and command smoothing, model-based feedback control, fuzzy and LMI-based control, data-driven and model-free adaptive control, reinforcement learning, and machine learning-based approaches. The control design is then developed from a compact Takagi–Sugeno fuzzy state-space representation of the nonlinear cart–pendulum crane dynamics. The fuzzy model is obtained by sector nonlinearity and local approximation in fuzzy partition spaces, resulting in two local rules scheduled by the load swing angle. A parallel distributed fuzzy controller is parameterized as a linear-in-the-parameters control law, and its local parameters are adapted online by a teacher-based normalized least mean squares (LMS) rule. The adaptive controller learns the control signal of a virtual supervisory fuzzy controller and then remains active as the controller in the test phase. Projection, leakage regularization, force saturation, and a smooth transfer from the supervisory signal to the learned LMS signal are included to obtain an implementation-oriented closed-loop formulation. The proposed approach is intended as a physically interpretable bridge between robust fuzzy crane modeling and data-driven adaptive control. The novelty is clarified as the integration of a compact fuzzy crane model, a supervisory teacher signal, normalized LMS learning, leakage/projection, input saturation, and command-governed anti-sway tuning in one simulation-oriented framework. The study is explicitly presented as a numerical benchmark; formal nonlinear stability certification and laboratory validation are identified as future work rather than as claims of the present contribution.
- Open Access
- 30 July 2026
- Published by MDPI AG in Chemengineering
- Vol. 10 (8) , 95
Inverse Fuzzy Model Control: A Data-Driven Learning Framework with Application to Thermal Process Control
Inverse fuzzy model control (IFMC) is an attractive strategy for regulating nonlinear thermal processes, where complex heat-transfer dynamics, disturbances, and actuator constraints challenge conventional control approaches. However, traditional direct and indirect inverse schemes often exhibit limited generalization and sensitivity to noise. This paper presents a data-driven inverse fuzzy control framework for temperature regulation in thermal systems, combining Takagi–Sugeno modeling, functional cancelation feedback, and auxiliary-input optimization. Gaussian antecedents and affine consequents are jointly identified via nonlinear least squares from input–output data generated by the theoretical PHE model. The same theoretical model is used as the numerical plant in all closed-loop simulations, whereas the forward and inverse fuzzy models are constructed only from the generated data and do not use the analytical PHE equations during identification or online inference. The learned inverse model estimates control actions that achieve the desired temperature trajectory while respecting actuator constraints. The method is evaluated in simulation on a plate heat exchanger (PHE) benchmark under tracking and disturbance-rejection scenarios. Results show accurate temperature regulation, smooth actuator behavior, and disturbance rejection in the tested scenarios. Compared with a numerically tuned PI baseline, the proposed approach reduces tracking RMSE by 27.5% and the standard deviation of the primary control current by 9.6% in the reported simulation scenario, indicating improved tracking and smoother primary actuation. In general, the proposed framework provides an interpretable and efficient solution for nonlinear thermal processes, with potential applications in energy systems, heat exchangers, and related thermal engineering technologies.
- 30 July 2026
- Published by MDPI AG in Chemengineering
- Vol. 10 (8) , 95
Open AccessInverse Fuzzy Model Control: A Data-Driven Learning Framework with Application to Thermal Process Control
Inverse fuzzy model control (IFMC) is an attractive strategy for regulating nonlinear thermal processes, where complex heat-transfer dynamics, disturbances, and actuator constraints challenge conventional control approaches. However, traditional direct and indirect inverse schemes often exhibit limited generalization and sensitivity to noise. This paper presents a data-driven inverse fuzzy control framework for temperature regulation in thermal systems, combining Takagi–Sugeno modeling, functional cancelation feedback, and auxiliary-input optimization. Gaussian antecedents and affine consequents are jointly identified via nonlinear least squares from input–output data generated by the theoretical PHE model. The same theoretical model is used as the numerical plant in all closed-loop simulations, whereas the forward and inverse fuzzy models are constructed only from the generated data and do not use the analytical PHE equations during identification or online inference. The learned inverse model estimates control actions that achieve the desired temperature trajectory while respecting actuator constraints. The method is evaluated in simulation on a plate heat exchanger (PHE) benchmark under tracking and disturbance-rejection scenarios. Results show accurate temperature regulation, smooth actuator behavior, and disturbance rejection in the tested scenarios. Compared with a numerically tuned PI baseline, the proposed approach reduces tracking RMSE by 27.5% and the standard deviation of the primary control current by 9.6% in the reported simulation scenario, indicating improved tracking and smoother primary actuation. In general, the proposed framework provides an interpretable and efficient solution for nonlinear thermal processes, with potential applications in energy systems, heat exchangers, and related thermal engineering technologies.
- Open Access
- 2 April 2026
- Published by Wiley in Steel Research International
- Vol. 97 (9) , 4779-4804
Development and Validation of a Next‐Generation Mechanistic Model of the Electric Arc Furnace
The paper presents a mechanistic model of an electric arc furnace (EAF), covering thermal, mass‐transfer, and chemical processes. It represents a next step in the field of EAF modeling, building on the basis of our previous work; however, it incorporates substantial enhancements, including newly developed modules for heat and mass transfer, melting geometry, radiation, slag behavior, chemical reactions, off‐gas dynamics, and arc–slag interactions. These improvements lead to inclusion of more EAF processes, their accurate representation, and higher calculation accuracy, thus resolving several limitations of the previous model. Furthermore, parameterization and validation of the model using EAF measurements have shown that the model prediction errors for bath temperatures are between 10°C and 20°C, the errors for bath and slag chemistry around 20%, while the off‐gas composition and temperature errors are below 20%. These measures show that the model describes all crucial mechanisms of the EAF with sufficient detail, and can thus provide a useful tool for either process simulation, monitoring, or optimization.
- 2 April 2026
- Published by Wiley in Steel Research International
- Vol. 97 (9) , 4779-4804
Open AccessDevelopment and Validation of a Next‐Generation Mechanistic Model of the Electric Arc Furnace
The paper presents a mechanistic model of an electric arc furnace (EAF), covering thermal, mass‐transfer, and chemical processes. It represents a next step in the field of EAF modeling, building on the basis of our previous work; however, it incorporates substantial enhancements, including newly developed modules for heat and mass transfer, melting geometry, radiation, slag behavior, chemical reactions, off‐gas dynamics, and arc–slag interactions. These improvements lead to inclusion of more EAF processes, their accurate representation, and higher calculation accuracy, thus resolving several limitations of the previous model. Furthermore, parameterization and validation of the model using EAF measurements have shown that the model prediction errors for bath temperatures are between 10°C and 20°C, the errors for bath and slag chemistry around 20%, while the off‐gas composition and temperature errors are below 20%. These measures show that the model describes all crucial mechanisms of the EAF with sufficient detail, and can thus provide a useful tool for either process simulation, monitoring, or optimization.
- Open Access
- 12 January 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 190
Evolving granular fuzzy control: Overview, case study on the chaotic Hénon map, and research outlook
- 12 January 2026
- Published by Elsevier in Applied Soft Computing
- Vol. 190
Open AccessEvolving granular fuzzy control: Overview, case study on the chaotic Hénon map, and research outlook
- Open Access
A Neuro-Fuzzy System for Interpretable Long-Term Stock Market Forecasting
In the complex landscape of multivariate time series forecasting, achieving both accuracy and interpretability remains a significant challenge. This paper introduces the Fuzzy Transformer (Fuzzformer), a novel recurrent neural network architecture combined with multi-head self-attention and fuzzy inference systems to analyze multivariate stock market data and conduct long-term time series forecasting. The method leverages LSTM networks and temporal attention to condense multivariate data into interpretable features suitable for fuzzy inference systems. The resulting architecture offers comparable forecasting performance to conventional models such as ARIMA and LSTM while providing meaningful information flow within the network. The method was examined on the real world stock market index S\&P500. Initial results show potential for interpretable forecasting and identify current performance tradeoffs, suggesting practical application in understanding and forecasting stock market behavior.
Open AccessA Neuro-Fuzzy System for Interpretable Long-Term Stock Market Forecasting
In the complex landscape of multivariate time series forecasting, achieving both accuracy and interpretability remains a significant challenge. This paper introduces the Fuzzy Transformer (Fuzzformer), a novel recurrent neural network architecture combined with multi-head self-attention and fuzzy inference systems to analyze multivariate stock market data and conduct long-term time series forecasting. The method leverages LSTM networks and temporal attention to condense multivariate data into interpretable features suitable for fuzzy inference systems. The resulting architecture offers comparable forecasting performance to conventional models such as ARIMA and LSTM while providing meaningful information flow within the network. The method was examined on the real world stock market index S\&P500. Initial results show potential for interpretable forecasting and identify current performance tradeoffs, suggesting practical application in understanding and forecasting stock market behavior.
- Open Access
Black-Box Time-Series Domain Adaptation via Cross-Prompt Foundation Models
The black-box domain adaptation (BBDA) topic is developed to address the privacy and security issues where only an application programming interface (API) of the source model is available for domain adaptations. Although the BBDA topic has attracted growing research attentions, existing works mostly target the vision applications and are not directly applicable to the time-series applications possessing unique spatio-temporal characteristics. In addition, none of existing approaches have explored the strength of foundation model for black box time-series domain adaptation (BBTSDA). This paper proposes a concept of Cross-Prompt Foundation Model (CPFM) for the BBTSDA problems. CPFM is constructed under a dual branch network structure where each branch is equipped with a unique prompt to capture different characteristics of data distributions. In the domain adaptation phase, the reconstruction learning phase in the prompt and input levels is developed. All of which are built upon a time-series foundation model to overcome the spatio-temporal dynamic. Our rigorous experiments substantiate the advantage of CPFM achieving improved results with noticeable margins from its competitors in three time-series datasets of different application domains.
Open AccessBlack-Box Time-Series Domain Adaptation via Cross-Prompt Foundation Models
The black-box domain adaptation (BBDA) topic is developed to address the privacy and security issues where only an application programming interface (API) of the source model is available for domain adaptations. Although the BBDA topic has attracted growing research attentions, existing works mostly target the vision applications and are not directly applicable to the time-series applications possessing unique spatio-temporal characteristics. In addition, none of existing approaches have explored the strength of foundation model for black box time-series domain adaptation (BBTSDA). This paper proposes a concept of Cross-Prompt Foundation Model (CPFM) for the BBTSDA problems. CPFM is constructed under a dual branch network structure where each branch is equipped with a unique prompt to capture different characteristics of data distributions. In the domain adaptation phase, the reconstruction learning phase in the prompt and input levels is developed. All of which are built upon a time-series foundation model to overcome the spatio-temporal dynamic. Our rigorous experiments substantiate the advantage of CPFM achieving improved results with noticeable margins from its competitors in three time-series datasets of different application domains.
- Open Access
- 22 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (10) , 3736-3746
Evolving Gaussian Systems as a Framework for Federated Regression Problems
In this paper, we present a novel federated learning framework to multivariate regression problems, termed Evolving Gaussian Federated Regression (eGauss+$_{FR}$). The need for a federated approach is due to the increasing problem of distributed acquisition of the data and protection for the rights of distributing these data. Regression problems are usually nonlinear and therefore strongly connected to the clustering to divide the data space into smaller sub-spaces where a linear approximation could be applied. Here we are faced with the main drawback of traditional clustering methods, where a predefined number of clusters is needed. In federated learning problems, where the data is commonly non-identically distributed between different sources or clients, this represents a significant challenge. This problem can be overcome by introducing an evolving approach which add and removes the clusters on-the-fly. The idea in our approach is to use the incremental c-regression or c-varieties clustering method to define the clusters which lie close to the lines and describe them with the centers and the covariance matrices. The clustering is done for each data source or client. Due to the restriction and protection of data sharing, only the centers and the covariance matrices of all clients are then transmitted to main server and merged together, which is here done in a way as proposed in eGauss+ method. From merged clusters the auxiliary points are generated, which than serve to approximate the function by using classical fuzzy models. Our proposed method was demonstrated on simple synthetic data, while synthetic and real-world datasets were used to test time complexity and scalability with the number of clients. The results demonstrate the benefits of evolving federated method, which results in high quality approximation of the function and can be easily extended to high-dimensional problems.
- 22 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 33 (10) , 3736-3746
Open AccessEvolving Gaussian Systems as a Framework for Federated Regression Problems
In this paper, we present a novel federated learning framework to multivariate regression problems, termed Evolving Gaussian Federated Regression (eGauss+$_{FR}$). The need for a federated approach is due to the increasing problem of distributed acquisition of the data and protection for the rights of distributing these data. Regression problems are usually nonlinear and therefore strongly connected to the clustering to divide the data space into smaller sub-spaces where a linear approximation could be applied. Here we are faced with the main drawback of traditional clustering methods, where a predefined number of clusters is needed. In federated learning problems, where the data is commonly non-identically distributed between different sources or clients, this represents a significant challenge. This problem can be overcome by introducing an evolving approach which add and removes the clusters on-the-fly. The idea in our approach is to use the incremental c-regression or c-varieties clustering method to define the clusters which lie close to the lines and describe them with the centers and the covariance matrices. The clustering is done for each data source or client. Due to the restriction and protection of data sharing, only the centers and the covariance matrices of all clients are then transmitted to main server and merged together, which is here done in a way as proposed in eGauss+ method. From merged clusters the auxiliary points are generated, which than serve to approximate the function by using classical fuzzy models. Our proposed method was demonstrated on simple synthetic data, while synthetic and real-world datasets were used to test time complexity and scalability with the number of clients. The results demonstrate the benefits of evolving federated method, which results in high quality approximation of the function and can be easily extended to high-dimensional problems.
- Open Access
Measures of Overlapping Multivariate Gaussian Clusters in Unsupervised Online Learning
In this paper, we propose a new measure for detecting overlap in multivariate Gaussian clusters. The aim of online learning from data streams is to create clustering, classification, or regression models that can adapt over time based on the conceptual drift of streaming data. In the case of clustering, this can result in a large number of clusters that may overlap and should be merged. Commonly used distribution dissimilarity measures are not adequate for determining overlapping clusters in the context of online learning from streaming data due to their inability to account for all shapes of clusters and their high computational demands. Our proposed dissimilarity measure is specifically designed to detect overlap rather than dissimilarity and can be computed faster compared to existing measures. Our method is several times faster than compared methods and is capable of detecting overlapping clusters while avoiding the merging of orthogonal clusters.
Open AccessMeasures of Overlapping Multivariate Gaussian Clusters in Unsupervised Online Learning
In this paper, we propose a new measure for detecting overlap in multivariate Gaussian clusters. The aim of online learning from data streams is to create clustering, classification, or regression models that can adapt over time based on the conceptual drift of streaming data. In the case of clustering, this can result in a large number of clusters that may overlap and should be merged. Commonly used distribution dissimilarity measures are not adequate for determining overlapping clusters in the context of online learning from streaming data due to their inability to account for all shapes of clusters and their high computational demands. Our proposed dissimilarity measure is specifically designed to detect overlap rather than dissimilarity and can be computed faster compared to existing measures. Our method is several times faster than compared methods and is capable of detecting overlapping clusters while avoiding the merging of orthogonal clusters.
- Open Access
Federated Learning based on Self-Evolving Gaussian Clustering
In this study, we present an Evolving Fuzzy System within the context of Federated Learning, which adapts dynamically with the addition of new clusters and therefore does not require the number of clusters to be selected apriori. Unlike traditional methods, Federated Learning allows models to be trained locally on clients' devices, sharing only the model parameters with a central server instead of the data. Our method, implemented using PyTorch, was tested on clustering and classification tasks. The results show that our approach outperforms established classification methods on several well-known UCI datasets. While computationally intensive due to overlap condition calculations, the proposed method demonstrates significant advantages in decentralized data processing.
Open AccessFederated Learning based on Self-Evolving Gaussian Clustering
In this study, we present an Evolving Fuzzy System within the context of Federated Learning, which adapts dynamically with the addition of new clusters and therefore does not require the number of clusters to be selected apriori. Unlike traditional methods, Federated Learning allows models to be trained locally on clients' devices, sharing only the model parameters with a central server instead of the data. Our method, implemented using PyTorch, was tested on clustering and classification tasks. The results show that our approach outperforms established classification methods on several well-known UCI datasets. While computationally intensive due to overlap condition calculations, the proposed method demonstrates significant advantages in decentralized data processing.