Results 171 to 180 of about 1,182 (205)

Parallel tangent methods with variable stepsize

2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No.04CH37541), 2005
The most widely used algorithm for training multilayer feedforward neural networks is backpropagation. Back-propagation is an iterative gradient descent algorithm. Since its appearance, various methods which modify the conventional BP have been created to improve its efficiency. One such algorithm which uses an adaptive learning rate is backpropagation
Yiannis G. Petalas, Michael N. Vrahatis
openaire   +1 more source

Stability of variable and random stepsize LMS

1997 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2002
The stability of variable stepsize LMS (VSLMS) algorithms with uncorrelated stationary Gaussian data is studied. It is found that when the stepsize is determined by the past data, the boundedness of the step size by the usual stability condition of fixed stepsize LMS is sufficient for the stability of VSLMS.
Saul B. Gelfand, Yongbin Wei, James V. Krogmeier   +2 more
openaire   +1 more source

The Convergence of Variable-Stepsize, Variable-Formula, Multistep Methods

SIAM Journal on Numerical Analysis, 1984
Die Verff. studieren das Verhalten von Mehrschrittverfahren auf variablen Gittern der Gestalt \[ y_{n+1}=\sum^{s_ n}_{i=0}a_{n,i}y_{n- i}+b_ nh_{n+1}y'\!_{n+1}+\sum^{r_ n}_{i=0}b_{n,i}h_{n- i}y'\!_{n-i} \] mit y'\({}_ i:=f(t_ i,y_ i)\). Die Stabilität des linearen Teils der Verfahren allein hat (unter passenden Voraussetzungen) die des gesamten ...
Crouzeix, M., Lisbona, F. J.
openaire   +2 more sources

ESIRK methods and variable stepsize

Applied Numerical Mathematics, 1998
Singly implicit Runge-Kutta methods (SIRK) have been introduced in order to reduce the implementation costs. Many of them, however, have the disadvantage that certain internal stages approximate the solution outside the integration interval. The idea of effective order permits to overcome this difficulty, but leads to a less straightforward ...
Butcher, J. C., Chen, D. J. L.
openaire   +2 more sources

Variable order and stepsize in general linear methods

Numerical Algorithms, 2019
zbMATH Open Web Interface contents unavailable due to conflicting licenses.
Saghir Ahmad, John C. Butcher, Winston L. Sweatman   +2 more
openaire   +1 more source

Time-transformations for reversible variable stepsize integration

Numerical Algorithms, 1998
zbMATH Open Web Interface contents unavailable due to conflicting licenses.
Stephen D. Bond, Benedict J. Leimkuhler
openaire   +2 more sources

A Time-Reversible Variable-Stepsize Integrator for Constrained Dynamics

SIAM Journal on Scientific Computing, 1999
The main results of the paper are the design and implementation of variable-stepsize methods for the time-discretization of constrained Euler-Lagrange equations faithful to symplectic and time-reversal symmetry structures of the equations. Symplectic structures are first explored because of the fact that classical equations of motion can be put in ...
Eric Barth, Benedict J. Leimkuhler, Sebastian Reich   +2 more
openaire   +1 more source

Variable stepsize in Cowell's method

BIT, 1989
The author proposes a formula which is intended to facilitate fourth order piecewise constant step size implementation of Cowell's method.
openaire   +1 more source

Supplement to Construction of Variable-Stepsize Multistep Formulas

Mathematics of Computation, 1986
Supplement to the author's paper in ibid. 47, 503-510 (1986; reviewed above).
openaire   +2 more sources

Reversible Long-Term Integration with Variable Stepsizes

SIAM Journal on Scientific Computing, 1997
The problem \(y'=f(y)\), \(y(0)=y_0\) is called \(\rho\)-reversible \((\rho\) is an invertible linear transformation in the phase space) if for all \(y\) the equation \(f(\rho y)= -\rho f(y)\) holds. Some Hamiltonian systems belong to this category.
Hairer, Ernst, Stoffer, Daniel
openaire   +1 more source