Quick search Find article
Quick search
Find article

Transient dynamics of on-line learning in two-layered neural networks

Michael Biehl, Peter Riegler and Christian Wöhler

Show affiliations


The dynamics of on-line learning in neural networks with continuous units is dominated by plateaux in the time dependence of the generalization error. Using tools from statistical mechanics, we show for a soft committee machine the existence of several fixed points of the dynamics of learning that give rise to complicated behaviour, such as cascade-like runs through different plateaux with a decreasing value of the corresponding generalization error. We find learning-rate-dependent phenomena, such as splitting and disappearing of fixed points of the equations of motion. The dependence of plateau lengths on the initial conditions is described analytically and simulations confirm the results.


PACS

07.05.Mh Neural networks, fuzzy logic, artificial intelligence

02.30.-f Function theory, analysis

05.10.-a Computational methods in statistical physics and nonlinear dynamics

MSC

62M45 Neural nets and related approaches

68Q32 Computational learning theory (See also 68T05)

82C32 Neural nets (See also 68T05, 91E40, 92B20)

Subjects

Mathematical physics

Computational physics

Instrumentation and measurement

Statistical physics and nonlinear systems

Dates

Issue 16 (21 August 1996)

Received 1 March 1996



View by subject




Export








Please login to access our web services, or create an account if you don't yet have one.

You must have cookies enabled in your web browser to be able to login.

Username
Password

Forgotten your password? Get a new one here.