Mercurial > octave-libgccjit

comparison scripts/optimization/fsolve.m @ 18857:7bbe3658c5ef

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
maint: Use "FIXME:" coding convention in m-files. * flipdim.m, prepad.m, rotdim.m, doc.m, strread.m, textread.m, krylov.m, colon.m, dump_prefs.m, fileattrib.m, getappdata.m, __xzip__.m, unpack.m, fsolve.m, axis.m, meshc.m, print.m, __ghostscript__.m, __go_draw_axes__.m, __print_parse_opts__.m, struct2hdl.m, unique.m, spstats.m, treeplot.m, test.m, datestr.m: Use "FIXME:" coding convention in m-files.
author Rik <rik@octave.org>
date Wed, 25 Jun 2014 13:45:41 -0700
parents d63878346099
children
comparison
equal deleted inserted replaced
18856:de8c67ba7ac4 18857:7bbe3658c5ef
380 ## is that for a tolerance that depends on scaling, only 0 makes 380 ## is that for a tolerance that depends on scaling, only 0 makes
381 ## sense as a default value. But 0 usually means uselessly long 381 ## sense as a default value. But 0 usually means uselessly long
382 ## iterations, so we need scaling-independent tolerances wherever 382 ## iterations, so we need scaling-independent tolerances wherever
383 ## possible. 383 ## possible.
384 384
385 ## FIXME -- why tolf*n*xn? If abs (e) ~ abs(x) * eps is a vector 385 ## FIXME: Why tolf*n*xn? If abs (e) ~ abs(x) * eps is a vector
386 ## of perturbations of x, then norm (fjac*e) <= eps*n*xn, i.e. by 386 ## of perturbations of x, then norm (fjac*e) <= eps*n*xn, i.e. by
387 ## tolf ~ eps we demand as much accuracy as we can expect. 387 ## tolf ~ eps we demand as much accuracy as we can expect.
388 if (fn <= tolf*n*xn) 388 if (fn <= tolf*n*xn)
389 info = 1; 389 info = 1;
390 ## The following tests done only after successful step. 390 ## The following tests done only after successful step.