datafit: disturbing unconsistencies and missing important informations PARAME
Reported by Samuel GOUGEON (@sgougeon)
-- Bug description --
datafit: disturbing unconsistencies and missing important informations
PARAMETERS:
* W: weighting matrice. Default is eye(ne,ne)
* algo: bounds are said to be set on x, while x is not defined (p instead).
A reference to http://forge.scilab.org/upload/docoptimscilab/files/optimization_in_scilab.pdf would be appreciated.
Remark: This useful document does not refer a single time to datafit()
* df0: refers to f, that is neither an arg nor defined. Having a look at help optim, we understand that f is the scalar output of the cost function. However, in datafit, the cost function is not straightforwardly accessible. The only somewhat related quantity is G, whose output is not scalar. Therefore, the definition of df0 is as puzzling as its use. In the description part, it would be worthwhile to explicitely define as "f" in the expression: "Vector p is found by minimizing f = G(p,z_1)...etc
* stop/nap: reference to undocumented "fun". G instead?
stop/epsf: reference to undocumented f (as above)
stop/epsx: reference to undocumented x (as above)
* "in": Reference to unknown fun (G instead?)
DESCRIPTION: Is datafit() based on optim()? (looks very likely).
REFERENCES: Reference to the PDF by Baudin et al., and others, would be welcome.
-- Scilab error message --
-- How to reproduce the bug --
help datafit