Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

BCCN/BFNT Sonderseminar

Friday, 04.12.2009 13 s.t.

Learning invariances via topographic maps in a network of spiking neurons

by Dr. Thomas Wachtler
from BCCN/G-Node Munich

Location

DPZ, Seminarraum des Cognitive Neuroscience Lab

Abstract

Our visual system recognizes objects fairly independent of viewing conditions, and it achieves a stable representation of visual space independent of eye movements. We developed a model network of spiking neurons that learns these invariances by combining temporal correlation learning with self-organizing map formation. When trained with stimulus statistics corresponding to natural viewing situations, the network is able to learn representations of visual objects invariant to changes in viewing angle, and to learn coordinate transformations from a retina-centered to a head-centered frame of reference that is invariant to changes in gaze direction. ----------------------------------------------------------------------- The German Neuroinformatics Node - Tool and Data Sharing for Neurophysiology The German Neuroinformatics Node, G-Node (www.gnode.org), has been established to facilitate the interaction and collaboration between experimental and theoretical neuroscientists, both nationally and internationally. G-Node is funded by the German Federal Ministry for Education and Research (BMBF) and is an integral part of the Bernstein Network for Computational Neuroscience (www.nncn.de). The current focus of G-Node is in the area of cellular and systems neurophysiology, developing infrastructure and tools for data analysis and data sharing in this field, together with teaching and training.

back to overview