Dynamic feature linking in stochastic networks with.pdfVIP

Dynamic feature linking in sto chastic networks with short range interactions Bert Kapp en 1 and Pablo Varona2 1 RW CP Novel Function SNN Lab oratory Department of Biophysics University of Nijmegen Geert Gro oteplein NL EZ Nijmegen The Netherlands 2 Departamento de Ingenieria Informatica Universidad Autonoma de Madrid Canto Blanco Madrid Spain Abstract It is well established that cortical neurons display synchronous ring for some stimuli and not for others The resulting synchronous subp opulation of neurons is thought to form the basis of ob ject p ercep tion In this pap er this dynamic linking phenomenon is demonstrated in networks of binary neurons with sto chastic dynamics Feedforward con nections implement feature detectors and lateral connections implement memory traces or cell assemblies Intro duction It is well established that cortical neurons display synchronous ring for some stimuli and not for others In particular it has b een shown exp erimentally that correlations dep end on the amount of conict in the stimulus presented The resulting synchronous subp opulations of neurons cell assemblies are thought to form the basis of segmentation and ob ject p erception The role of individual cells is to represent imp ortant atomic visual features such as edges corners velo cities colors etc Ob jects can b e dened as a collection of these features The cell assembly is a neural representation of the entire ob ject Thus if the lo cal features are part of a coherent global stimulus the corresp onding neurons synchronize If the same lo cal features are not part of a global stimulus no such synchronization o ccurs The observed sy