The 'moving targets' training algorithm

Richard Rohwer; D. S. Touretzky

The 'moving targets' training algorithm

Richard Rohwer, D. S. Touretzky (Editor)

Research output: Unpublished contribution to conference › Unpublished Conference Paper › peer-review

Abstract

A simple method for training the dynamical behavior of a neural network is derived. It is applicable to any training problem in discrete-time networks with arbitrary feedback. The algorithm resembles back-propagation in that an error function is minimized using a gradient-based method, but the optimization is carried out in the hidden part of state space either instead of, or in addition to weight space. Computational results are presented for some simple dynamical training problems, one of which requires response to a signal 100 time steps in the past.

Original language	English
Publication status	Unpublished - 1990
Event	Advances in Neural Information Processing Systems 1990 - Dublin, United Kingdom Duration: 29 Aug 1990 → 31 Aug 1990

Other

Other	Advances in Neural Information Processing Systems 1990
Country/Territory	United Kingdom
City	Dublin
Period	29/08/90 → 31/08/90

Keywords

dynamical behavior
neural network
error

Access to Document

misc90-003.pdf

Cite this

@conference{545ddb27106b43e3b0c7e0d57ed2bc56,

title = "The 'moving targets' training algorithm",

abstract = "A simple method for training the dynamical behavior of a neural network is derived. It is applicable to any training problem in discrete-time networks with arbitrary feedback. The algorithm resembles back-propagation in that an error function is minimized using a gradient-based method, but the optimization is carried out in the hidden part of state space either instead of, or in addition to weight space. Computational results are presented for some simple dynamical training problems, one of which requires response to a signal 100 time steps in the past.",

keywords = "dynamical behavior, neural network, error",

author = "Richard Rohwer and Touretzky, {D. S.}",

year = "1990",

language = "English",

note = "Advances in Neural Information Processing Systems 1990 ; Conference date: 29-08-1990 Through 31-08-1990",

}

TY - CONF

T1 - The 'moving targets' training algorithm

AU - Rohwer, Richard

A2 - Touretzky, D. S.

PY - 1990

Y1 - 1990

N2 - A simple method for training the dynamical behavior of a neural network is derived. It is applicable to any training problem in discrete-time networks with arbitrary feedback. The algorithm resembles back-propagation in that an error function is minimized using a gradient-based method, but the optimization is carried out in the hidden part of state space either instead of, or in addition to weight space. Computational results are presented for some simple dynamical training problems, one of which requires response to a signal 100 time steps in the past.

AB - A simple method for training the dynamical behavior of a neural network is derived. It is applicable to any training problem in discrete-time networks with arbitrary feedback. The algorithm resembles back-propagation in that an error function is minimized using a gradient-based method, but the optimization is carried out in the hidden part of state space either instead of, or in addition to weight space. Computational results are presented for some simple dynamical training problems, one of which requires response to a signal 100 time steps in the past.

KW - dynamical behavior

KW - neural network

KW - error

M3 - Unpublished Conference Paper

T2 - Advances in Neural Information Processing Systems 1990

Y2 - 29 August 1990 through 31 August 1990

ER -

The 'moving targets' training algorithm

Abstract

Other

Keywords

Access to Document

Fingerprint

Cite this