supplierasebo.blogg.se

1. #INCOMPLETE GAUSS NEWTON INVERSION RES2DINV PC#
2. #INCOMPLETE GAUSS NEWTON INVERSION RES2DINV DOWNLOAD#

The function call is designed to allow use on both (weakly) nonlinear and linear problems, both regularized (inverse) and non-regularized (parameter estimation) problems, and both frequentist and Bayesian problems, and using a user-supplied function for the Jacobian matrix of derivatives or internally calculating finite differences, as determined by the form of its input parameters. Perhaps in future versions INVGN will allow these options to be set via keyword/value pairs in the input argument list, but not implemented yet. InvGN is compatible with Octave (version >3), but to handle a few minor remaining differences between Matlab and Octave, do note the "usingOctave" flag among the options listed at top of the InvGN.m script.

PE Gill, W Murray, & MH Wright, "Practical Optimization," Academic Press, 1981.

#INCOMPLETE GAUSS NEWTON INVERSION RES2DINV PC#

RC Aster, B Borchers, & CH Thurber, "Parameter Estimation and Inverse Problems," Elsevier Academic Press, 2004.Professor Ken Creager's ESS-523 inverse theory class, Univ of Washington, 2005.Hansen 1987), and the L-curve is one of InvGN's outputs. The damping level (regularization parameter lambda) is solved for by the L-curve method (see e.g. Where epsilon is some model norm threshold. Where delta is some statistically-determined noise threshold, and also to: The damped nonlinear least squares problem is:įor appropriate choices of regularization parameter lambda, this problem is equivalent to:

#INCOMPLETE GAUSS NEWTON INVERSION RES2DINV DOWNLOAD#

(Browse the contents and download the package file for InvGN at my GitHub account!)