direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Page Content

All Publications

Unspecific LTP May Lead to Simultaneous and Robust ON-OFF and Eye-Dominance Segregation in Correlation Based Learning Models
Citation key Stetter1998c
Author Stetter, M. and Lang, E.W., and Weih, P. and Obermayer, K.
Title of Book Proceedings
Year 1998
Publisher Machines that Learn Workshop, Snowbird
Abstract It has been shown that activity-driven segregation of elongated ON- and OFF-dominated patches within simple receptive fields [2,6] or the development of cortical ocular dominance patterns [3] can be desribed in the framework of correlation-based Hebbian learning. Simultaneous OD and ON-OFF segregation, however, does not occur without additional assumptions such as specific nonlinearities, a \\\"competitiveintracortical dynamics, or a specific LGN circuitry [4,5]. \\ In a recent study performed in hippocampal slices, Engert and Bonhoeffer [1] reported unspecific LTP for synapses which are located less than 100 micrometer away from each other. Here we report that under certain conditions on the distribution of synapses on the dendrites of cortical cells, this kind of unspecific LTP leads to a robust, simultaneous segregation of ON-OFF and OD, even for positive between-eye correlations. Our model consists of a network with four layers of LGN neurons (left eye ON-center, left eye OFF-center, right eye ON-center, right eye OFF-center) and one layer which corresponds to layer 4C in primary visual cortex. LGN neurons are characterized by Mexican hat shaped receptive fields and project retinotopically to the cortical layer, which is equipped with a Mexican-hat shaped lateral interaction function. Transfer functions are piecewise linear with a rectifying and a saturating nonlinearity. The learning rule consists of a traditional Hebbian and an unspecific term. Analytical results for the case of linear transfer functions show that the unspecific component may lead to simultaneous OD and ON-OFF segregation, if same-eye synapses form groups on their postsynaptic cell\'s dendritic tree. These results are confirmed by numerical simulations for neurons with linear as well as rectifying and saturating transfer functions. Our results are also in accord with experiments on strabismic cats. The study implies that details of synaptic plasticity can considerably influence the predictions of high-level developmental models. \\ References: \\ [1] F. Engert and T. Bonhoeffer (1997), Nature 388, 279-284.\\ [2] K. D. Miller (1994), J. Neurosci. 14, 409-441. \\ [3] K. D. Miller (1989), Science 245, 605-615. \\ [4] K. Obermayer, H. Ritter, and K. Schulten (1990), Proc. Natl. Acad. Sci. USA 87, 8345-8349. \\ [5] C. Piepenbrock, H. Ritter and K. Obermayer (1996), Neur. Proc. Lett. 3, 31-37. \\ [6] M. Stetter, A. Muller, and E. Lang (1994),Phys. Rev. E 50, 4167-4181.
Download Bibtex entry

Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe

Auxiliary Functions