Title :
A non-quadratic gradient algorithm
Author_Institution :
Dept. of Electr. Eng., Virginia Univ., Charlottesville, VA, USA
Abstract :
Based on a non-quadratic cost function, a gradient-type algorithm is developed for estimating parameters of a linear model, resulting in a bounded normalized estimation error which converges to zero asymptotically with time and belongs to L1+α for 0<α<1. Preliminary analysis and simulation results show that the new algorithm leads to a faster convergence of the parameter error to small values than the standard gradient algorithm with α=1
Keywords :
convergence of numerical methods; error analysis; linear systems; optimisation; parameter estimation; estimation error; linear model; non-quadratic cost function; non-quadratic gradient algorithm; parameter error; Algorithm design and analysis; Analytical models; Convergence; Cost function; Ear; Estimation error; Parameter estimation; Signal design; Vectors;
Conference_Titel :
Decision and Control, 1994., Proceedings of the 33rd IEEE Conference on
Conference_Location :
Lake Buena Vista, FL
Print_ISBN :
0-7803-1968-0
DOI :
10.1109/CDC.1994.411710
