Thesis supervisor: Gergő Bognár
Location of studies (in Hungarian): ELTE IK Numerikus Analízis Tanszék Abbreviation of location of studies: ELTE
Description of the research topic:
Two predominant approaches in the field of signal and image processing are the model-based approach that relies on mathematical, statistical, or physical models, and the data-driven approach based on machine learning using artificial intelligence. The research aims to develop model-based artificial intelligence methods that combine both approaches and support the direct incorporation of domain knowledge into machine learning algorithms. Modell-based architectures promises higher efficiency, reliability, and explainability. The research objective is to investigate adaptive transformation methods (adaptive projection in particular), adaptive filters, dynamical system models, and feature extraction techniques; their combination with model-based machine learning approaches (primarily with model-based neural networks); architecture development and optimization; and application to real-world signal and image processing problems (e.g. filtering, reconstruction, segmentation, classification of natural images or medical signals and images).
Required language skills: angol Further requirements: Fundamental knowledge of mathematical modelling, signal and image processing, and machine learning; programming and software development skills in Python and Matlab environments.