Projects and Grants

Team members focus on the development of techniques for the efficiency of language models and their integration into a comprehensive system. This system will enable researchers in the humanities and other non-technical fields to perform advanced analyses of text corpora. Among the techniques studied are weight quantization, distillation of large models into smaller versions, and the creation of ensembles of expert models that are dynamically activated based on the characteristics of the input data. Another area of research is the identification and detection of disinformation using a range of methods based on artificial intelligence, machine learning, and large language models. These methods include, for example, network spread analysis, which can detect unusual increases in certain narratives, and the study of network structures in the disinformation space. Time series analysis is also used to detect anomalies, which allows for the identification of unusual patterns in the spread of information.

Team members in this broader project provide mathematical and informatics support for research on topics related to the energy transition in the Ostrava metropolitan area. In addition to developing our own methods, we also collaborate with other teams, including the development of a mathematical model for PEDs (Positive Energy Districts) and data analysis of large datasets to answer questions related to the study of energy poverty.

The main goal of the team’s involvement in this broad strategic project focused on regional transformation is to conduct research on current trends in data analysis and the development of methods for artificial intelligence applications. These methods are designed for use across a wide range of fields, including social science research and industrial applications. The objective is not only to develop data analytics methods but also to apply them in collaboration with other work packages and research labs.

Team members focus on research, development, and implementation of computational intelligence methods for the analysis of uncertain information, such as biosensor and social science data with systematic errors. In the analytical part, they focus on hypothesis generation, non-statistical processing of time series with detection and prediction of structural changes, and machine learning, including deep neural networks. Research on knowledge systems aims at expert knowledge bases using Explainable AI (XAI) principles in combination with data-driven and compositional models. This is crucial for personalized digital interventions and mHealth, where data-driven approaches are insufficient to determine the appropriate intervention, which requires expert knowledge and logical methods of interpretation.