AG Kommunikationstheorie

Thema:

Investigating lateral information processing in the olfactory system using optimization theory

Abstract:

A non-negative least squares (NNLS) optimization problem is a convex optimization problem, which, in contrary to a standard least squares problem, does not have an analytical solution. Instead, there exist efficient iterative algorithms to find the optimal solution. In this talk, a new approach is presented. A specially designed non-linear and discontinuous dynamic system is introduced. The system is shown to converge to the solution of the NNLS problem and to be Lyapunov stable. By an appropriate modelling, the neural network in the olfactory system implements this dynamic system and shows good performance in the tasks of odour decomposition and detection. Thus, the presented concept is an interesting new approach for the purposes of lateral information processing in the olfactory system. The talk presents an extract of my masterthesis.



zurück zur Terminübersicht