Implementation of the Gauss-Newton Algorithm for Non-linear Least-mean-squares Fitting in FPGA Devices

Abstract-- The paper presents the implementation of nonlinear least-squares regression in a Field Programmable Gate Array (FPGA) device. The implemented algorithm is very performant in obtaining the coefficients of nonlinear functions from a set of input data using the least squares regression using the method of Gauss-Newton. The convergence time of the algorithm is greatly smaller with respect to microprocessors and DSPs at state of art with a power dissipation well below 10W.

• Type: Journal
• Year: 2009
• Where: ERSA
• Authors: Andrea Abba, Antonio Manenti, Andrea Suardi, Angelo Geraci, Giancarlo Ripamonti
• Keywords: ERSA 2009, Field Programmable Gate Array, Hardware, Implemented Algorithm, Squares Regression