From: Luis Ibanez <luis.ibanez@ki...>  20020223 23:57:22

Hi I'm using the optimizers in vnl and couldn't find so far a way of getting access to the set of parameters that are the final solution to the optimization. So I'm now writing cost functions that store the set of parameters of the last call to f() or gradf() and assuming that when the optimizers returns its last evaluation was the one corresponding to the optimal set of parameters. Is there a way of getting these parameters directly from the optimizer ? Thanks for any help Luis 
From: Amitha Perera <perera@cs...>  20020224 05:16:37

When you call the minimize function of a minimizer, the input parameter is typically an initial point, and will be modified to be the minimizing point: vnl_conjugate_gradient cg; ... vnl_vector x = (starting point) cg.minimize( x ); vcl_cout << "Minimum value has parameters " << x << vcl_endl; Is this what you meant? Amitha. On Sat, Feb 23, 2002 at 06:57:20PM 0500, Luis Ibanez wrote: > Hi > > I'm using the optimizers in vnl and couldn't find so > far a way of getting access to the set of parameters > that are the final solution to the optimization. > > So I'm now writing cost functions that store the > set of parameters of the last call to f() or gradf() > and assuming that when the optimizers returns its > last evaluation was the one corresponding to the > optimal set of parameters. > > Is there a way of getting these parameters directly > from the optimizer ? > > > Thanks for any help > > > Luis > > > > > _______________________________________________ > Vxlusers mailing list > Vxlusers@... > https://lists.sourceforge.net/lists/listinfo/vxlusers 
From: Luis Ibanez <luis.ibanez@ki...>  20020224 05:20:44

Yes, that's what I was looking for. Thanks Luis  Amitha Perera wrote: >When you call the minimize function of a minimizer, the input >parameter is typically an initial point, and will be modified to be >the minimizing point: > > vnl_conjugate_gradient cg; > ... > vnl_vector x = (starting point) > cg.minimize( x ); > vcl_cout << "Minimum value has parameters " << x << vcl_endl; > >Is this what you meant? > >Amitha. > > >On Sat, Feb 23, 2002 at 06:57:20PM 0500, Luis Ibanez wrote: > >>Hi >> >>I'm using the optimizers in vnl and couldn't find so >>far a way of getting access to the set of parameters >>that are the final solution to the optimization. >> >>So I'm now writing cost functions that store the >>set of parameters of the last call to f() or gradf() >>and assuming that when the optimizers returns its >>last evaluation was the one corresponding to the >>optimal set of parameters. >> >>Is there a way of getting these parameters directly >>from the optimizer ? >> >> >> Thanks for any help >> >> >> Luis >> >> >> >> >>_______________________________________________ >>Vxlusers mailing list >>Vxlusers@... >>https://lists.sourceforge.net/lists/listinfo/vxlusers >> > 
From: Andrew Fitzgibbon <awf@ro...>  20020224 15:27:04

The vector of parameters which you supplied as the initial guess is changed after the routine returns to hold the final answer. It is generally not true that the last call to f() was at the lowest value. 