In the era of rapid advancements in signal-capturing and computing, exploring brain activities has reached new heights through innovative machine learning-based methodologies. This takes us beyond conventional limitations, inviting us to step from uncertainty into the realm of understanding.
Ensemble clustering and diversity problem
M.Sc. Reza Mahini’s dissertation presents the development of various ensemble clustering methods for exploring interesting brain patterns. Cluster examination of brain signals emerges as a pivotal tool, delving into the stability of brain activity and testing the similarity of samples.
The explored stable pattern from the brain more likely corresponds to a specific brain response to environmental or designed stimulation or natural brain activity. In this, even advanced methods performance can be varied for different contexts. This problem is well known as the "elephant in the dark room” story or diversity problem in science.
Clustering for brain insights and biomarkers
Combining insights from various clustering methods ensures a reliable solution, requiring minimal prior knowledge about the intricacies of brain activity.
“The goal here is to automatically find interesting patterns in the signals from our brains — patterns connected to how we think, as well as patterns connected to important brain jobs like seeing, hearing and paying attention”, explains Mahini.
This research not only addresses the diversity problem in science but also paves the way for a deeper comprehension of the mind's complexities, offering a new context for the characterization of brain biomarkers.
M.Sc. Reza Mahini defends his doctoral dissertation “Consensus clustering for group-level analysis of event-related potential data” on 7.12.2023 at noon in Agora, room Ag B103 or online via Zoom. Opponent is Professor Raju S. Bapi (International Institute of Information Technology, India) and custos is Professor Timo Hämäläinen (Ģֱ). The language of the dissertation is English.
Publishing information
The dissertation “Consensus clustering for group-level analysis of event-related potential data” can be read on the JYX publication archive: ;
