Mailman 3 February 2007 - SciPy-User

Does Scipy have the differenciation function?
by Hui Chang Moon 24 Feb '07

24 Feb '07

Hello, Scipy-user Group members, I want to know if the Scipy has the differencitaion function. I can find the integration fuction (quad), but I can't know where the differenciation function. Whoever knows the differenciation function, let me know. Thank you.

3 2

[CFP] Zope conference on Zope in Science
by Andreas Jung 23 Feb '07

23 Feb '07

Dear Python & Zope Community, the eighth Zope conference organized by the German Zope User Group (DZUG) will be held this year at the Potsdam Institute for Climate Impace Research from 4. to 5 June 2007 (near Berlin). The topic of the conference will be Zope in sciences. Proposals for talks and workshops can be submitted until 01.04.2007 http://www.zope.de/redaktion/dzug/tagung/potsdam-2007/dzug-conference-2007-… or http://www.zope.de/redaktion/dzug/tagung/potsdam-2007/dzug-conference-2007-… Both German and English proposals are highly welcome. You can find further information about the Zope conference here: http://www.zope.de/8-dzug-tagung Regards, Andreas Jung Assistant Chairman DZUG e.V.

1 0

error in std?
by oliver.tomic＠matforsk.no 23 Feb '07

23 Feb '07

Hi list, we are in the process of switching our application from: (OLD configuration) Python 2.4.2 scipy 0.49 numpy 0.98 to: (NEW configuration) Python 2.5 scipy 0.52 numpy 1.0.1 When I do the following in the OLD configuration: data = array([1,2,3,4,5,6]) stand = std(data) the result for stand is 1.870828693386..... ,which is exactly what should be. However, when I do exactly the same under the NEW configuration the result for stand is 1.70782512766. I am clueless. Has anyone experienced the same problem? Any help appreciated. Oliver

3 3

Clipborad more friendly tha IDE
by Giorgio Luciano 23 Feb '07

23 Feb '07

Sorry for the easy question, I've searched the web but find nothing "really" helpful. Is there some editor more user friendly for Windows to have a workspace where to copy and paste from gnumeric/excel like the matlab workspace I'm using orange, which is wonderful, but if you have a lot of array to copy and paste is not so comfortable as a simple command line x=[(ctrl+v)] Thanks in advance to all. I guess that a workspace like matlab for numpy/scipy will make a lot of people to switch more easily form matlab :))) Giorgio

2 1

Mac OS X 10.4 can not import from scipy.linalg
by Jason Addison 23 Feb '07

23 Feb '07

I installed the prebuilts: MacPython python-2.4.4-macosx2006-10-18.dmg MacPython.mpkg SciPy ScipySuperpack-Intel-10.4-py2.4.dmg PyMC-1.1-py2.4-macosx10.4.mpkg gfortranCompleteInstaller.mpkg matplotlib-0.87.7-py2.4-macosx10.4.mpkg numpy-1.0.2.dev3522-py2.4-macosx10.4.mpkg scipy-0.5.3.dev2630-py2.4-macosx10.4.mpkg I'm using a MacBook Pro with Mac OS 10.4.8 with fink and Mac Developer Tools installed. After installing, I tried the tutorial: jra$ python Python 2.4.4 (#1, Oct 18 2006, 10:34:39) [GCC 4.0.1 (Apple Computer, Inc. build 5341)] on darwin Type "help", "copyright", "credits" or "license" for more information. >>> from numpy import matrix >>> from scipy.linalg import inv, det, eig Traceback (most recent call last): File "<stdin>", line 1, in ? File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/linalg/__init__.py", line 8, in ? from basic import * File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/linalg/basic.py", line 17, in ? from lapack import get_lapack_funcs File "/Library/Frameworks/Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/linalg/lapack.py", line 17, in ? from scipy.linalg import flapack ImportError: Inappropriate file type for dynamic loading >>> Searching for possibly relevent conflicting files I find stuff like: lapack: clapack.h in vecLib.framework blas: blas.hpp in ublas from boost numeric Does anyone have any idea on how to fix this? Is it supposed to work? Do I have something else installed that is conflicting? Thanks for your help ... jra

1 0

io.read_array with strings
by Charles-Antoine Robelin 22 Feb '07

22 Feb '07

I have been using io.read_array successfully to read ASCII files containing integers and floats, and I would like to import strings into arrays as well. I tried with io.read_array, but did not get it to work: If I create an array manually (i.e., numpy.array([['a1', 'd3', 'gg'],['wq', 'ty', 'e']])), the type (dtype) of its elements is '|S4', so I suspect numpy.arrays can handle strings. However, io.read_array(<filename>, separator=',') on the following file: a1, d3, gg wq, ty, e returns an array of floats with the correct shape, containing the numbers it could find (a1 -> 1.; d3 -> 3.) and 0. where no number could be found. I could not find how to force the type "strings," such as atype='<strings>' in the call of read_array. Is importing strings possible with io.read_array, or with another function, without having to parse manually? Thanks in advance.

2 1

Fwd: [sage-devel] Fwd: (Summer Of Code) are you interested in numerical optimization (+solving non-smooth non-linear systems of equations)?
by Fernando Perez 22 Feb '07

22 Feb '07

[ Forwarded from the SAGE dev list, since this may well be right up someone's alley from the scipy crowd] ---------- Forwarded message ---------- From: William Stein <wstein(a)gmail.com> Date: Feb 22, 2007 1:00 PM Subject: [sage-devel] Fwd: (Summer Of Code) are you interested in numerical optimization (+solving non-smooth non-linear systems of equations)? To: sage-devel(a)googlegroups.com, suvrit(a)cs.utexas.edu Does anybody have any thoughts about this potentially excellent idea of a package that could help SAGE? ---------- Forwarded message ---------- From: dmitrey <openopt(a)ukr.net> Date: Feb 22, 2007 1:56 AM Subject: (Summer Of Code) are you interested in numerical optimization (+solving non-smooth non-linear systems of equations)? To: wstein(a)gmail.com Hallo William Stein! I found your email address in http://wiki.python.org/moin/SummerOfCode/Mentors As far as I understood there are very few constrained solvers for Python. In google I failed to find anything but the CVXOPT, which consists mostly of wrappers to commercial mosek, and some LP/MIP wrappers to GNU C- or f- code (and some optimization routines from scipy, of course). So I'm last-year post-graduate (institute of cybernetics, Ukraine national science academy, optimization department). Our department research methods of optimization for non-smooth (& noisy) funcs since 1964 or so (under leadership of academician Naum Z. Shor till 2002 when he gone away), and parallel department under leadership of academician I. Sergienko & dr. V.Shilo researches combinatory optimization problems (something like matlab bintprog, MAXCUT etc, some weeks ago they published article of their GRASP-based code that won comparison vs CPLEX & some other commercial solvers). So all our software is opensourse & free, mostly fortran & C written. 3-4 months ago I began to write (in m-files) OpenOpt for MATLAB/Octave (1st ver 0.15 had been reliesed in November 25, 2006). It's equivalent to commercial TOMLAB or GAMS (currently puny of course, but free (GNU GPL2)) and contains 4 global solvers, that I connected to the OpenOpt environment from matworks fileexchange, 2 my local nonsmooth solvers - ShorEllipsoid for nVars = 1...10 & ralg for nVars = 1...1000, and nonSmoothSolve - MATLAB fsolve equivalent for non-smooth & noisy funcs. There is a good comparison in Examples/nonSmoothSolveEx.m, which shows, that fsolve fails even on low-noise or low-non-smooth funcs, but nonSmoothSolve - not. ralg & ShorEllipsoid are MATLAB equivalents to fminsearch; however, they can handle (as long as nonSmoothSolve): lb<=x<=ub; Ax<=b; Aeq*x=beq; c(x)<=0; h(x)=0; %as far as I understood from CVXOPT documentation it can't handle these constraints - see http://www.ee.ucla.edu/~vandenbe/cvxopt/doc/e-nlcp.html as long as gradients or subgradients df, dc, dh of f, c, h. MATLAB fminsearch can't handle anything mentioned above, and in 95% it fails comparison to ralg (but I can't say I tried too much examples). Just try, for example, f(x)=sum(abs(x).^1.2.^(0:(length(x)-1))); x0 = cos(1:60).'; or Lemarechal.m from OpenOpt/test (convex, continuous, non-smooth) These and some more examples are in OpenOpt/Examples, see ooexample5.m, ooexample2.m and others. This directory also contains some pictures of convergence, automatically generated by the files. Also OpenOpt performs auto scaling (but I not tested properly it yet); providing patterns of f, c, h (when no (sub)gradient is provided by user) can greatly speedup calculations; possibility of parallel calculations while obtaining df/dx numerically (via MATLAB dfeval, Octave users must provide similar func in prob.parallel.fun) and some more features. So are you interested in Python version of the OpenOpt? If yes, I probably would be able to contact with other Kiev institute of System Analis, where a group under leadership of academician Pshenichniy (perished some months ago) & dr. Nikitin develop smooth algorithms of optimization, and their IP-based smooth solvers (constrained, of course; including 2nd-order solvers) are considered to be one of essential. So I probably would be able to write for you Pyhton OpenOpt ver (GNU GPL2) with essential equivalent of MATLAB fmincon, as long as ralg, ShorEllipsoid, bintprog; drawing pictures of convergence, using patterns of dependences, parallel obtaining of (sub)gradients to f(), c(), h(); network problems solvers etc. If you have any questions, or can add any financial support, or want to look at my CV (I had about 1-1.5 years of Python experience & 3-4 years of optimization in MATLAB), or anything else - you can contact me via email or icq 275 - 976 - 670 (invisible). The more financial support can be obtained, the more time can I spend for the OpenOpt Python version, & the more icyb optim department workers can be involved into the OpenOpt for Python development (any salary amount are essential to the Ukraine workers). If I gain enough money, I propose: creating the same environment for Python as it is done for MATLAB/Octave; (1-1.5 months) writing ralg() & ShorEllipsoid() solvers (Unconstrained: ~1 week, constrained: +2-3 weeks) writing nonSmoothSolve() : ~ 1-2 weeks writing MATLAB bintprog (f*x->min, A*x<=b, Aeq*x=beq) equivalent based on rd. Shilo (& others) version of GRASP (works better than current CPLEX!): ~2 weeks writing MATLAB fmincon equivalent (smooth constrained optimization, c(x)<=0, h(x)=0, linear constraints +1st & 2nd derivatives) based on works of dr. Nikitin & academician Pshenichniy: (I can't estimate time for now, I must contact to him before) -------- + following implementation of other solvers, their upgrade & maintenance. I guess in future you could easily connect the py-code to the SAGE project, for example like you did with MAXIMA package. some links: ftp where OpenOpt versions are storing: http://www.box.net/public/6bsuq765t4 (you would better to download OpenOpt0.36.tar.bz2 from here) my page at matlabexchange area http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=13… Let me attach a graphic file generated by the OpenOpt ooexample3.m WBR, Dmitrey -- William Stein Associate Professor of Mathematics University of Washington --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel(a)googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscribe(a)googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/ -~----------~----~----~----~------~----~------~--~---

1 0

nan bug in distributions.norm.cdf
by Christopher Fonnesbeck 22 Feb '07

22 Feb '07

For some reason, perfectly valid normal random variates return a nan when passed to the normal cdf in the stats.distributions package: In [10]: from scipy.stats import distributions as d ... In [31]: d.norm.cdf(-0.73646593092) Out[31]: array(nan) In [32]: d.norm.cdf(-0.7) Out[32]: array(0.24196365222307303) In [33]: d.norm.cdf(-0.8) Out[33]: array(0.21185539858339669) Simply rounding this value makes it work. Not sure why this happens. Using a relatively recent svn build on OSX. cf

3 3

ANN: PyDSTool now compatible with numpy 1.0.1, scipy 0.5.2 and 64-bit CPUs.
by Rob Clewley 21 Feb '07

21 Feb '07

We are pleased to announce version 0.84 of PyDSTool, an open-source dynamical systems simulation, modeling, and analysis package. This long-overdue release is primarily intended to bring existing PyDSTool functionality up to date with the latest numpy and scipy releases (previous versions required scipy 0.3.2, numarray, numeric, etc). Also, PyDSTool is now compatible with 64-bit CPUs. While we have added a few new features and made several fixes, major improvements to functionality are in the pipeline for version 0.90. Please see http://pydstool.sourceforge.net for release notes and documentation, and http://sourceforge.net/projects/pydstool for downloading. As ever, please send us feedback if you have any problems with this new release or ideas and code contributions for future releases. Regards, Rob, Erik, and Drew. Center for Applied Mathematics, Cornell University. ****************** PyDSTool is an integrated simulation, modeling and analysis package for dynamical systems, written in Python (and partly in C). It is being developed at Cornell University, and the source code is available under the terms of the BSD license. PyDSTool runs on Linux, Windows, and Macs, and aims to have a minimal number of package dependencies.

3 5

Sort and N dimensional array by 1d array
by Christopher Clarke 21 Feb '07

21 Feb '07

Hi I am looking for an efficient way put a two dimensional array in the same order as a 1 d array In the case of arrays key, b and c key=array([1,3,6,0 ,4]) b = array([6,7,8,9,11]) c=array([[6,10],[7,11],[8,6],[2,5],[21,55]]) ixs=argsort(key) sb=take(b,ixs) however for when i do sc=take(c,ixs) sc.shape=(5,) (i've lost the second column) does this mean i have to go take(c[:,0],ixs) take(c[:,1],ixs) and concatenate the results. I know i am supposed to figure out how to specify the 2d sorted indecies for the 2 d array but??? Regards Chris

2 2