In his doctoral dissertation Mehmet Yamac addresses challenges which have historically limited AI's application, such as neural networks' black-box and data-hungry nature, their instability, lack of interpretability, and the absence of privacy-conscious data processing, alongside their high computational demands. He presents hybrid modeling via integration of sparse presentation and AI data representation and harmonizes robust mathematical frameworks with AI’s learning-driven approach. This approach addresses these long-standing challenges in data processing and introduces a novel level of efficiency and interpretability to AI algorithms. The synergy enhances decision-making, diagnostic accuracy, and sustainable technology development, even amidst high computational demands.
Among the notable application areas of the mathematical framework in his dissertation is the development of AI-driven models capable of personalized anomaly detection and privacy-preserving monitoring systems. These models, designed with a focus on interpretability and efficiency pave the way for potential applications providing technologically advanced and socially impactful solutions. The doctoral research also has the potential to address technology's environmental impact, presenting AI designs with a significantly enhanced carbon handprint.
Public Defence on 19 March
The doctoral dissertation of Mehmet Yamac, in the field of signal processing and machine learning entitled Advances in sparse representation: efficient modeling and applications will be publicly examined at the Faculty of Information Technology and Communication Sciences of Tampere University in room TB109 of Tietotalo Building (address: Korkeakoulunkatu 1, Tampere) at 12:00 on Tuesday 19 March 2024.
The Opponent will be by Professor Gonzalo R. Arce from the University of Delaware, USA. The Custos will be Professor Moncef Gabbouj from Tampere University.