Identifying Neuron Assemblies in Massively Parallel Spike Trains
15.08.2012 - 14:15 - 15:00
C222, Exactum, Kumpula
Dr. Christian Borgelt, Principal Researcher of European Centre for Soft Computing, Mieres (Asturias), Spain, will give a guest lecture on Identifying Neuron Assemblies in Massively Parallel Spike Trains.
D.O. Hebb (1949) suggested that cell assemblies form the building blocks of information processing in the brain. The member neurons are assumed to show correlated activity, corresponding to the temporal coincidence hypothesis about how groups of neurons encode stimuli and generally information. In this talk I present two approaches that are able to detect neuron assemblies in massively parallel spike data both reliably and efficiently. The first method focuses on whether individual neurons participate in correlated spiking activity. I present a family of specialized statistics and a surrogate data based testing procedure for this task. The second method relies on frequent item set mining algorithms and extends early work by G. Gerstein et al. (1978) on the accretion algorithm, which is designed to detect groups of neurons that exhibit coincident firing activity. By studying the likelihood of random occurrences of patterns having a certain number of neurons and a certain number of coincidences with the help of surrogate data generation, a statistically sound and still fairly efficient procedure to detect neuron assemblies reliably is obtained.
Christian Borgelt obtained his Ph.D. in 2000 from the University of Magdeburg, Germany, with a thesis on "Data Mining with Graphical Models" and the venia legendi for computer science in 2006 with a thesis on "Prototype-based Classification and Clustering." Since 2006 he is a principal researcher at the European Center for Soft Computing in Mieres, Spain, where he leads the Intelligent Data Analysis and Graphical Models Research Unit. His research interests include various data mining and machine learning methods like decision and regression trees, association rules, graphical models, frequent graph mining, clustering algorithms, neural networks and computational geometry. The current focus of his research is on adapting frequent pattern mining algorithms to specific tasks in application areas like neuroscience and traffic data analysis.