[SciPy-User] Passing fmin_cg a gradient

[Robert Kern]

On Fri, Apr 15, 2011 at 16:26, Jason Rennie <jrennie at gmail.com> wrote:
> On Wed, Apr 13, 2011 at 3:41 PM, Christian K. <ckkart at hoc.net> wrote:
>>
>> I have been working on the same thing a couple of years ago during my
>> PhD. In the end I found, that fmin_tnc is performing best among all
>> minimizers available in scipy.
>
> I'm actually on the look-out for a good open source CG or L-BFGS
> implementation.  Thanks for the fmin_tnc mention---I'll check it out.  Do
> you know what the "tnc" refers to and/or what are the differences with the
> other implementations?

http://js2007.free.fr/code/index.html#TNC

"""
TNC is a C implementation of TNBC, a truncated newton optimization
package originally developed by Stephen G. Nash in Fortran.
The original source code can be found at Stephen G. Nash Software Page.
This software aims at minimizing the value of a nonlinear function
whose variables are subject to bound constraints. It requires to be
able to evaluate the function and its gradient and is especially
useful for solving large scale problems.
The C source code and an example, along with license and copyright
information are included in the package. This software has been tested
on the majority of UNIXes and on Windows.
Since version 1.0.5, a Python interface module is provided. This
interface can also be found in SciPy.
"""

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless
enigma that is made terrible by our own mad attempt to interpret it as
though it had an underlying truth."
  -- Umberto Eco