[Scipy-tickets] [SciPy] #1510: Add Hessian method to Scipy

SciPy Trac scipy-tickets@scipy....
Mon Sep 5 17:53:40 CDT 2011


#1510: Add Hessian method to Scipy
---------------------------------------------------------------+------------
 Reporter:  deil                                               |       Owner:  somebody   
     Type:  enhancement                                        |      Status:  new        
 Priority:  normal                                             |   Milestone:  Unscheduled
Component:  scipy.optimize                                     |     Version:  0.9.0      
 Keywords:  Hesse, Hessian, numerical differentiationrical di  |  
---------------------------------------------------------------+------------
 It would be nice if scipy included numerical differentiation methods.
 I'm not sure where it would fit best, maybe in scipy.optimize or in
 scipy.misc?

 scipy.optimize already contains an approx_fprime method, which
 approximates
 the Jacobian. Having a second method approx_hessian to compute the
 Jacobian
 would be very useful, because the covariance matrix can be computed as the
 inverse of the Hessian, and having this in scipy would make it possible to
 compute
 fit parameter errors for other optimizers than scipy.optimize.leastsq.

 There are already two numpy-based python packages that implement various
 variants of  numerical differentiation routines:

 http://code.google.com/p/numdifftools/

 https://github.com/statsmodels/statsmodels/blob/master/scikits/statsmodels/sandbox/regression/numdiff.py

 I am not enough familiar with the differences in the methods implemented
 to make a suggestion which methods would be most useful to add to scipy.
 Maybe the authors can comment?

 The need for Hessians came up in this mailing list discussion:

 http://mail.scipy.org/pipermail/scipy-user/2011-September/030459.html

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1510>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list