User talk:Petergans/b/sandbox

Linear and nonlinear

 * Should perhaps underline that linearity/nonlinearity is with respect to the fitting parameters, rather than the underlying variables.
 * In particular it should perhaps be emphasised that a function
 * $$y = \Sigma \, a_i \phi_i (\mathbf{x})$$
 * although nonlinear in x, is linear in the weights ai, so can still be fitted by Linear least squares.


 * On the other hand NLLS will be required when the function looks like
 * $$y = \psi(\mathbf{a}, \mathbf{x})$$
 * -- Jheald (talk) 23:04, 30 January 2008 (UTC)

Marquandt parameter

 * Should mention Moré approach for setting it, used in the best production codes.
 * Perhaps mention connection to trust-region, with λ coming out as a Lagrange multiplier. Jheald (talk) 23:04, 30 January 2008 (UTC)

Non-linear or nonlinear?
Should there be a dash in the article name or not? I sort of like it with the dash, the way Petergans wrote it, but I guess both are fine. Oleg Alexandrov (talk) 04:14, 1 February 2008 (UTC)


 * I prefer without the hyphen-dash, following Nonlinear system. Is there some sort of guideline for this in the math style guide? CRGreathouse (t | c) 20:21, 1 February 2008 (UTC)