Chair of Artificial Intelligence

The Chair of Artificial Intelligence conducts research on new methods of data mining, also such as lack sufficient precision of description, which take into account incomplete cases and various causes of incompleteness. We develop methods of multi-strategic modeling of knowledge, its arithmetic and interrelational analysis of dependencies. We create unique solutions in the field of biomedical image processing and analysis algorithms which support effective diagnosis of malignant melanoma.


Research teams
As part of the research topic „Next-generation tools for federated learning effectively supporting decision support systems with uncertainty”, a research group has been established in the Department of Artificial Intelligence with the main goal of developing methods and tools in federated learning systems taking into account imprecision, uncertainty and data gaps. To this end, it is planned to use multiple data sources to teach effective artificial intelligence models without sharing source data (federated learning with data security and anonymity) taking into account imprecision.
Federated learning (FL) is a particular approach to training machine learning (ML) algorithms in a way that means the data remains private. Specifically, federated learning techniques aim to train machine learning algorithms on multiple distributed devices or servers, each with its own local and private data.
This collaborative approach contrasts with traditional machine learning techniques, which are centralized and rely on collecting all data samples in one unique data set before using them. It also differs from parallel computing-based techniques, which were developed to optimize ML on multiple processors, using a centralized data set that is divided into identically distributed subsets for computation.
The goal of the research is to improve the efficiency of systems supporting federated learning using methods that take into account data uncertainty, guaranteeing member privacy while resisting attacks and ensuring fairness between system members.
In many areas of life, such as industry, medicine or economics, a problem arises when an organization does not have a large enough data set to construct a decision/diagnostic system of sufficient quality. In this case, data from different organizations must be used, which is related to the problem of data sharing. To overcome these problems, federated learning is becoming increasingly popular, enabling automatic learning across distributed networks of autonomous partners without sharing raw data.
An important goal of the emerging team is to strengthen the indicated research area, through cooperation with the University of Rzeszow and the internationally leading Maastricht University.
Team composition:
Barbara Pękala, PhD, DSc., Assoc. Prof.
prof. Anna Wilbik (Maastricht University)
Teresa Mroczek PhD. (UITM)
Dorota Gil MSc. (UITM)
Ewa Rak PhD. (UR)
Jarosław Szkoła MSc. (UR)
Dawid Kosior MSc. (UR)
Piotr Grochowalski PhD. (UR)

Scientific research projects financed from external sources:

Digital solutions for automatic skin cancer diagnosis
Project manager for the Polish team: Prof. J. W. Grzymała-Busse, for the German team: Prof. Jens Haueisen
The aim of the project is to develop methods and algorithms for image processing that will support the diagnosis of malignant melanoma performed by dermatologists and general practitioners, providing objective and reliable results. It will be implemented in cooperation with Technische Universität Ilmenau and the JensLab company.
Period of implementation: 05.2019–04.2022
Financing: National Center for Research and Development (NCBiR)

Scientific research projects financed from the Ministry:

Exploration of incomplete data and biomedical imaging
Project manager: Teresa Mroczek, Ph.D
As part of this research, two separate goals are proposed: 1) creating new methods for analyzing incomplete data sets and moving away from the commonly used approach in incomplete data mining, i.e. imputing missing attribute values before the mining process for the development of incompleteness interpretation methods and set indicators, and 2) development of new methods enabling quantitative assessment of biomedical images, including mainly the development of a method for automatic classification of digital images of melanocytic skin
nevi using deep learning networks.
Period of implementation: 2023-2025

Hybrid models using fuzzy and approximate set theories and their application in medicine
Project menager: Barbara Pękala, Ph.D., Assoc. Prof.
Development of Rough-Fuzzy models for image processing and analysis in the problem of “Optimization of physiotherapy offices” and development of an optimal hybrid rule system based on expert knowledge and machine learning.
Period of implementation: 2023-2024

Models of fuzzy and rough sets theory used for posture detection
Project leader: Barbara Pękala, Ph.D., Assoc. Prof.
The aim of the research is to develop new methods of computational intelligence constructed for uncertain data requiring non-classical methods of representation and modelling. In particular, sets of fuzzy and approximate inference algorithms for detecting body posture will be identified, which demonstrate effectiveness at a given level of accuracy and quality of individual classification methods.
Implementation period: 2021-2023

University of Information Technology and Management in Rzeszow

ul. Sucharskiego 2,
35-225 Rzeszow, Poland

phone: 17 866 11 11
fax: +48 17 866 12 22

All rights reserved, University of Information Technology and Management in Rzeszow © 2008-2023