From 4ac5c193e4502e5191dccf55fd325f4ccbb5eebe Mon Sep 17 00:00:00 2001 From: Davis King Date: Thu, 2 Dec 2010 23:49:07 +0000 Subject: [PATCH] Replaced a funny character with the normal ASCII one. --HG-- extra : convert_revision : svn%3Afdd8eb12-d10e-0410-9acb-85c331704f74/trunk%403935 --- dlib/optimization/optimization_least_squares_abstract.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/dlib/optimization/optimization_least_squares_abstract.h b/dlib/optimization/optimization_least_squares_abstract.h index 4f7273c2d..c407890a7 100644 --- a/dlib/optimization/optimization_least_squares_abstract.h +++ b/dlib/optimization/optimization_least_squares_abstract.h @@ -44,7 +44,7 @@ namespace dlib - This function performs an unconstrained minimization of the least squares function g(x) defined by: - g(x) = sum over all i: 0.5*pow( f(list(i),x), 2 ) - - This method combines the Levenberg–Marquardt method with a quasi-newton method + - This method combines the Levenberg-Marquardt method with a quasi-newton method for approximating the second order terms of the hessian and is appropriate for large residual problems (i.e. problems where the f() function isn't driven to 0). In particular, it uses the method of Dennis, Gay, and Welsch as described in @@ -92,7 +92,7 @@ namespace dlib - This function performs an unconstrained minimization of the least squares function g(x) defined by: - g(x) = sum over all i: 0.5*pow( f(list(i),x), 2 ) - - This method implements a plain Levenberg–Marquardt approach for approximating + - This method implements a plain Levenberg-Marquardt approach for approximating the hessian of g(). Therefore, it is most appropriate for small residual problems (i.e. problems where f() goes to 0 at the solution). - Since this is a trust region algorithm, the radius parameter defines the initial