Mailman 3 December 2006 - SciPy-User

C-extensions for NumPy code available.
by Lou Pecora 08 Dec '06

08 Dec '06

For those of you who do numerical calculations, need to speed up your code, and use NumPy arrays: I've written several C extensions that handle NumPy arrays. They are simple, but they seem to work well. They will show you how to pass Python variables and NumPy arrays to your C code. Once you learn how to do it, it's pretty straight-forward. I suspect they will suffice for most numerical code. I've written it up as a draft and have made the code and document file available. If you want a copy, just email me or post a followup here to this message. I will send it to you as a tar.gzip file (only 192 KB). Remember it is only a draft (no guarantees or warantees) and you should test the code for your own uses. I would really love people who are much more knowledgable about these things than I to look it over and set me straight on any mistakes (like not INREF'ing something). In addition, is there a place I could put this up on the web for others? I don't have any way to do that at my lab. I would like to share with the Python community since I've been helped so much on these mailing lists. By the way I did this after a long time of searching for ways to speed up Python and connect with my C and C++ code. There are a lot of solutions out there, but I decided that they all provided so much more than I needed and required so much more learning time that it wasn't worth it. I am not knocking them. I found for my numerical needs I really only need to pass a limited set of things (integers, floats, strings, and NumPy arrays). If that's your category, this code might help you. -- Lou Pecora Naval Research Lab Washington, DC ____________________________________________________________________________________ Need a quick answer? Get one in minutes from people who know. Ask your question on www.Answers.yahoo.com

3 4

help on numerical modelisation of near-sinusoidal function using SciPy for pendulum analysis
by Robert VERGNES 08 Dec '06

08 Dec '06

Hello, new to Python. I would like to know if in SciPy there is a already a fucntion for the modelisation of a near sinusoidal function. ie: I have result from data acquisition from a pendulum and I obtain a text file with a matrix: (t(sec), Theta(deg)). And I 'know' that is it is in the form of Theta(t) = a+b*sin(co*t+phi) And I want to give my matrix [t,Theta] to a Python Lib class and get in return the nearest parameters a,b,co and phi ( and if possible the error level ). Does it exist in Python librairy or SciPy or Pynum??? Currently I am doing that job with Regressi.exe from evariste - it works ok. but I would like to enclose this calculation in a larger set of other analysis hence I would like to get a function doing that modelisation - regression from the curve to the function we believe right. Thanx Robert --------------------------------- Yahoo! Mail réinvente le mail ! Découvrez le nouveau Yahoo! Mail et son interface révolutionnaire. --------------------------------- Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses.

2 2

Definition of random variable
by Jonas Kahn 08 Dec '06

08 Dec '06

Hi I try to create a new instance of the rv_continuous class, and the __init__ is not very explicit... Essentially, I would like to define it from the probability distribution only, and I guess (totally without any ground for that) that it is implemented as such. Especially I would like to use the associated inverse survival function, without writing it by myself. Of course I could use the routines for solving equations to get it... Alternatively, is there any way to get the inverse function of a monotonic function? Thanks for the help Jonas PS: I think there is something wrong with the "scale" parameter in the package, or I have not understood how to use it. There is no change to the pdf when I try to give it another value as 1: In [124]: stats.binom.rvs(1,1) Out[124]: array([1]) # No problem with loc In [126]: stats.binom.rvs(1,1, loc = 10) Out[126]: array([11]) # But nothing with scale, with or without loc In [127]: stats.binom.rvs(1,1, loc = 10, scale =100) Out[127]: array([11]) In [128]: stats.binom.rvs(1,1, scale =100) Out[128]: array([1])

1 1

Re: [SciPy-user] help on numerical modelisation of near-sinusoidal function using SciPy for pendulum analysis
by palazzol＠comcast.net 07 Dec '06

07 Dec '06

Hi, I think optimize.minpack.leastsq() could be made to work, but it would be best if you could make good initial guesses for a,b,co, and phi. You could use some ad-hoc methods to come up with guesses for these initial conditions, based on looking at a bit of the waveform, and then use leastsq() to calculate the best values and residual error norm. For example, a is the average over time, b is the half the amplitude between peaks, co based on Theta(t) crossing a, etc. You could even give it analytical Jacobians if you want to take some derivatives of your equation. Another alternative would be to take an fft(), and look at the DC offset, gain, maximum peak freq and phase to get good initial conditions. Good luck, -Frank -------------- Original message ---------------------- From: Robert VERGNES <robert.vergnes(a)yahoo.fr> > > Hello, > > new to Python. > I would like to know if in SciPy there is a already a fucntion for the > modelisation of a near sinusoidal function. ie: > I have result from data acquisition from a pendulum and I obtain a text file > with a matrix: > (t(sec), Theta(deg)). > > And I 'know' that is it is in the form of > > Theta(t) = a+b*sin(co*t+phi) > > And I want to give my matrix [t,Theta] to a Python Lib class and get in return > the nearest parameters a,b,co and phi ( and if possible the error level ). > > Does it exist in Python librairy or SciPy or Pynum??? > > Currently I am doing that job with Regressi.exe from evariste - it works ok. but > I would like to enclose this calculation in a larger set of other analysis hence > I would like to get a function doing that modelisation - regression from the > curve to the function we believe right. > > Thanx > > Robert > > > > --------------------------------- > Yahoo! Mail réinvente le mail ! Découvrez le nouveau Yahoo! Mail et son > interface révolutionnaire. > > > --------------------------------- > Découvrez une nouvelle façon d'obtenir des réponses à toutes vos questions ! > Profitez des connaissances, des opinions et des expériences des internautes sur Yahoo! Questions/Réponses.

3 2

Building SciPy on OS X: Universal binary issues
by Josh Marshall 07 Dec '06

07 Dec '06

I've been running for a while with older versions of numpy and scipy, but I now need to provide a Universal Binary version of my app for distribution. I'm running a G4 with 10.4.8. I have installed the universal versions of Python, Numpy, SciPy, matplotlib, etc from http://pythonmac.org/packages/py24-fat/ I get an error when importing scipy, as follows: ImportError: Failure linking new module: /Library/Frameworks/ Python.framework/Versions/2.4/lib/python2.4/site-packages/scipy/ special/_cephes.so: Symbol not found: _printf$LDBLStub Referenced from: /Library/Frameworks/Python.framework/Versions/2.4/ lib/python2.4/site-packages/scipy/special/_cephes.so Expected in: dynamic lookup My understanding is that this is a common error when a package is built with GCC 4, but this shouldn't crop up in a downloadable package. Where are instructions for building SciPy as a universal binary? Obviously it is being done, but I would like to know how to do it myself. The wiki would be a good spot for this to go. As a second issue, running the following command: for FILE in `find . | grep "\.so"` ; do file $FILE ; done (which gives the architecture for all the dynamic libraries) shows that only a few of the libraries are Universal. The output is attached below. ./cluster/_vq.so: Mach-O fat file with 2 architectures ./cluster/_vq.so (for architecture ppc): Mach-O bundle ppc ./cluster/_vq.so (for architecture i386): Mach-O bundle i386 ./fftpack/_fftpack.so: Mach-O bundle ppc ./fftpack/convolve.so: Mach-O bundle ppc ./integrate/_odepack.so: Mach-O bundle ppc ./integrate/_quadpack.so: Mach-O bundle ppc ./integrate/vode.so: Mach-O bundle ppc ./interpolate/_fitpack.so: Mach-O bundle ppc ./interpolate/dfitpack.so: Mach-O bundle ppc ./io/numpyio.so: Mach-O fat file with 2 architectures ./io/numpyio.so (for architecture ppc): Mach-O bundle ppc ./io/numpyio.so (for architecture i386): Mach-O bundle i386 ./lib/blas/cblas.so: Mach-O fat file with 2 architectures ./lib/blas/cblas.so (for architecture ppc): Mach-O bundle ppc ./lib/blas/cblas.so (for architecture i386): Mach-O bundle i386 ./lib/blas/fblas.so: Mach-O bundle ppc ./lib/lapack/atlas_version.so: Mach-O fat file with 2 architectures ./lib/lapack/atlas_version.so (for architecture ppc): Mach-O bundle ppc ./lib/lapack/atlas_version.so (for architecture i386): Mach-O bundle i386 ./lib/lapack/calc_lwork.so: Mach-O bundle ppc ./lib/lapack/clapack.so: Mach-O fat file with 2 architectures ./lib/lapack/clapack.so (for architecture ppc): Mach-O bundle ppc ./lib/lapack/clapack.so (for architecture i386): Mach-O bundle i386 ./lib/lapack/flapack.so: Mach-O fat file with 2 architectures ./lib/lapack/flapack.so (for architecture i386): Mach-O bundle i386 ./lib/lapack/flapack.so (for architecture ppc): Mach-O bundle ppc ./linalg/_flinalg.so: Mach-O bundle ppc ./linalg/_iterative.so: Mach-O bundle ppc ./linalg/atlas_version.so: Mach-O fat file with 2 architectures ./linalg/atlas_version.so (for architecture ppc): Mach-O bundle ppc ./linalg/atlas_version.so (for architecture i386): Mach-O bundle i386 ./linalg/calc_lwork.so: Mach-O bundle ppc ./linalg/cblas.so: Mach-O fat file with 2 architectures ./linalg/cblas.so (for architecture ppc): Mach-O bundle ppc ./linalg/cblas.so (for architecture i386): Mach-O bundle i386 ./linalg/clapack.so: Mach-O fat file with 2 architectures ./linalg/clapack.so (for architecture ppc): Mach-O bundle ppc ./linalg/clapack.so (for architecture i386): Mach-O bundle i386 ./linalg/fblas.so: Mach-O bundle ppc ./linalg/flapack.so: Mach-O bundle ppc ./linsolve/_csuperlu.so: Mach-O fat file with 2 architectures ./linsolve/_csuperlu.so (for architecture ppc): Mach-O bundle ppc ./linsolve/_csuperlu.so (for architecture i386): Mach-O bundle i386 ./linsolve/_dsuperlu.so: Mach-O fat file with 2 architectures ./linsolve/_dsuperlu.so (for architecture ppc): Mach-O bundle ppc ./linsolve/_dsuperlu.so (for architecture i386): Mach-O bundle i386 ./linsolve/_ssuperlu.so: Mach-O fat file with 2 architectures ./linsolve/_ssuperlu.so (for architecture ppc): Mach-O bundle ppc ./linsolve/_ssuperlu.so (for architecture i386): Mach-O bundle i386 ./linsolve/_zsuperlu.so: Mach-O fat file with 2 architectures ./linsolve/_zsuperlu.so (for architecture ppc): Mach-O bundle ppc ./linsolve/_zsuperlu.so (for architecture i386): Mach-O bundle i386 ./ndimage/_nd_image.so: Mach-O fat file with 2 architectures ./ndimage/_nd_image.so (for architecture ppc): Mach-O bundle ppc ./ndimage/_nd_image.so (for architecture i386): Mach-O bundle i386 ./optimize/_cobyla.so: Mach-O bundle ppc ./optimize/_lbfgsb.so: Mach-O bundle ppc ./optimize/_minpack.so: Mach-O bundle ppc ./optimize/_zeros.so: Mach-O fat file with 2 architectures ./optimize/_zeros.so (for architecture ppc): Mach-O bundle ppc ./optimize/_zeros.so (for architecture i386): Mach-O bundle i386 ./optimize/minpack2.so: Mach-O bundle ppc ./optimize/moduleTNC.so: Mach-O fat file with 2 architectures ./optimize/moduleTNC.so (for architecture ppc): Mach-O bundle ppc ./optimize/moduleTNC.so (for architecture i386): Mach-O bundle i386 ./signal/sigtools.so: Mach-O fat file with 2 architectures ./signal/sigtools.so (for architecture ppc): Mach-O bundle ppc ./signal/sigtools.so (for architecture i386): Mach-O bundle i386 ./signal/spline.so: Mach-O fat file with 2 architectures ./signal/spline.so (for architecture ppc): Mach-O bundle ppc ./signal/spline.so (for architecture i386): Mach-O bundle i386 ./sparse/sparsetools.so: Mach-O bundle ppc ./special/_cephes.so: Mach-O bundle ppc ./special/specfun.so: Mach-O bundle ppc ./stats/futil.so: Mach-O bundle ppc ./stats/mvn.so: Mach-O bundle ppc ./stats/statlib.so: Mach-O bundle ppc ./weave/examples/fibonacci_ext.so: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV), not stripped ./weave/examples/ramp_ext.so: ELF 32-bit LSB shared object, Intel 80386, version 1 (SYSV), not stripped

1 0

where I can find quadratic programming in scipy?
by Xiaojian Wang 07 Dec '06

07 Dec '06

Hi, I try to look for a module currently, is there any scipy module for quadratic programming? if not, anybody knows where I can get it free in any other languages? thanks Xiaojian

1 0

A Call to Arms - Advocating Python in Science
by Jeff Rush 07 Dec '06

07 Dec '06

Greetings. I've been lurking on this list, to check out what is needed to support Python in the scientific community. I'm the Python Advocacy Coordinator that the PSF has hired to help out. My main jumping off point for advocacy topics is at: http://wiki.python.org/moin/Advocacy One of the items I'm hoping to get input about is the contents of two advocacy kits I'm getting together: - Research Lab Python Advocacy Kit - University Educator's Python Advocacy Kit Details at: http://wiki.python.org/moin/Advocacy#AdvocacyKits I'm drumming up a series of whitepapers, flyers, videos and podcasts to put into those kits. Not being a scientist myself, I need ideas on what is needed. The actual individual items are being solicited at: http://wiki.python.org/moin/AdvocacyWritingTasks http://wiki.python.org/moin/ArticleIdeas. Next, we are seeking reusable/retargetable teaching materials, such as those under a Creative Commons license. We need slide presentations and class handouts. Now I know there are a great many slide presentations on the web about Python. I can google them all but we're not looking for just any presentation, we're looking for the best of field. We're collecting links and ideas at: http://wiki.python.org/moin/Advocacy#TeachingMaterials I know some of this already exists in the scientific community and I just need some nudges to come up to speed on what is available for me to make use of. Also permissions for reuse and reformatting, in some cases. Thanks for any assistance you have time to provide, Jeff Rush <advocate(a)python.org> Python Advocacy Coordinator

1 0

A comparison of Matlab's and SciPy's eig function for repeated eigenvalues
by Nils Wagner 07 Dec '06

07 Dec '06

Hi all, The eigenvectors associated with repeated eigenvalues are not uniquely defined. The eigenvectors returned by MATLAB's eig function are orthogonal with respect to A and B [V,D] = eig(A,B) V' * A * V = diag(a_i) V' * B * V = diag(b_i) And the eigenvalues differs with respect to the order. How can I obtain "similar" results with SciPy ? w, vr = linalg.eig(A,B) dot(vr.T,dot(A,vr)) \ne diag(i_i) dot(vr.T,dot(B,vr)) \ne diag(b_i) I have attached small test scripts for both, SciPy and MATLAB. So, what is the difference between MATLAB's eig and SciPy's eig ? Thanks in advance. Nils %%MatrixMarket matrix array real symmetric % 5 5 2.0000000000000000e+01 -1.0000000000000000e+01 0.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 2.0000000000000000e+02 -4.0000000000000000e+01 -1.0000000000000000e+02 -5.0000000000000000e+01 4.0000000000000000e+01 0.0000000000000000e+00 0.0000000000000000e+00 1.0000000000000000e+02 0.0000000000000000e+00 5.0000000000000000e+01 %%MatrixMarket matrix array real symmetric % 5 5 2.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 5.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 4.0000000000000000e+00 0.0000000000000000e+00 0.0000000000000000e+00 1.0000000000000000e+01 0.0000000000000000e+00 5.0000000000000000e+00

1 0

Re: [SciPy-user] feed-forward neural network for python
by Marek Wojciechowski 07 Dec '06

07 Dec '06

On Wed, 06 Dec 2006 18:00:08 -0000, <scipy-user-request(a)scipy.org> wrote: > Marek Wojciechowski wrote: >> Hi! >> I released feed-forward neural network for python project at sourceforge >> (ffnet). I'm announcing it here because it depends on numpy/scipy >> tandem, >> so you, folks, are potential users/testers >> >> If anyone is interested please visit ffnet.sourceforge.net (and then >> post >> comments if any...) > Hi Marek, > thank you for you ANN implementation. I'm currently interested in > recurrent > neural networks so I hope you'll add them too in near future > Anyway, for those interested in ANN you can check conx.py module of > PyRobotics project. It's just one python file (182Kb of python source!) > inside a much bigger project. Conx.py is self contained and unfortunately > it needs Numeric and not numpy :(( > What about mixing ffnet and conx.py? Comapring to conx.py ffnet is much, much faster and much, much simpler. Faster thanks to scipy optimization routines and simpler because simplicity was the basic assumption when I created ffnet. To prove it: I trained a XOR network (2-2-1) in conx with: from pyrobot.brain import conx net=conx.Network() net.addThreeLayers(2,2,1) net.setInputs([[0., 0.], [0., 1.], [1., 0.], [1., 1.]]) net.setTargets([[1.], [0.], [0.], [1.]]) net.setLearning(0.5) net.setMomentum(0.8) net.setTolerance(0.001) net.train() This performed 5000 iterations in 34.6 s (and the tolerance has not been reached) With ffnet it can be done with: from ffnet import ffnet, mlgraph conec = mlgraph((2,2,1)) net = ffnet(conec) input=[[0., 0.], [0., 1.], [1., 0.], [1., 1.]]; target=[[1.], [0.], [0.], [1.]] net.train_momentum(input, target, eta=0.5, momentum=0.8, maxiter=5000) The above trains nework in 17.9 ms, and the fitness is perfect. ffnet is almost 2000 times faster than conx in this example! An the code is simpler. However there are some nice testing methods in conx, which can be used in ffnet. Thanks for this tip. -- Marek Wojciechowski

2 1

scipy wiki
by fred 06 Dec '06

06 Dec '06

Hi all, The new is UI is much better (& faster) ! Great job ! Anyway, I get an error when I try to access to http://scipy.org/Cookbook/MayaVi or http://scipy.org/Cookbook/MayaVi/Installation (and only these two wiki pages): ERROR: The installed docutils version is 0.3.7; version 0.3.10 or later is required. Cheers, -- http://scipy.org/FredericPetit

2 3