Influence of the Learning Gain on the Dynamics of Oja's Neural Network

摘要

In this paper, the dynamical behavior of Oja's neural netowrk is analyzed. Oja's net has been traditionally studied in the continuous-time context via some simplification procedures, some of them concerning the asymptotic behavior of the learning gain. The contribution of the paper is the study of a deterministic discrete-time (DDT) version, preserving the discrete-time form of the original network and allowing a more realistic behavior of the learning gain. As a consequence, the discrete-time nature of the new model leads to results which are drastically different to the ones known for the continuous-time formulation. Simulation examples support the presented results.