The Bernstein Center Freiburg

Informal Seminar
Andreas Leppert

Institute of Physics, University of Freiburg

The behavior of recurrent neural networks near critical points

In order to get a deeper insight into learning processes, we will address recurrent neural networks based on several types of random graphs. The networks operate in a parameter area that exhibits an interesting bifurcation scenario of its attractor states. We will explain its characteristics and there will be given an overview in how far the bifurcation behavior can be controlled by parameters that play a role in learning processes. This are for example globally weighted links, inhibitory neurons and assortativity.
Wednesday, June 12, 2013

14:15 h
BCF Library
First Floor
Hansastr. 9a
The talk is open to the public. Guests are cordially invited!