Mailman 3 August 2004 - SciPy-User

xSYMM from BLAS-3 in scipy?
by Fernando Perez Aug. 26, 2004

Aug. 26, 2004

Hi all, I'm running a pretty recent CVS of scipy, and I'm needing to use the blas3 xSYMM routines (dSYMM, specifically). I notice they are not there, and after poking around a bit, it would _seem_ like extending the xGEMM wrapper shouldn't be too hard. But since I'm not at all familiar with this code, I'm a bit afraid of breaking too many things, and right now (with scipy'04 coming up and 2 talks yet to be written :) I don't have the time to write the necessary tests to make this a committable change. Is there anyone out there who has wrapped the rest of the blas3 routines? A quick scan (by tab-completion and source study) seems to show that the only blas3 stuff wrapped is xGEMM. At least having xSYMM, xHEMM and xTRMM would be great, though I guess at some point we should try to have all of blas3. I can try to do this at a later time, but it would be great if someone already has this kind of code available. Best, f

1 0

Modular toolkit for Data Processing (MDP) released
by Pietro Berkes and Tiziano Zito Aug. 24, 2004

Aug. 24, 2004

We are pleased to announce the first public release of the MDP library for Python (http://mdp-toolkit.sourceforge.net). This package has been developed in the context of computational neuroscience research, but it should fit the needs of a larger audience of scientists and developers. Modular toolkit for Data Processing is a Python library to implement data processing elements (nodes) and to combine them into data processing sequences (flows). A node corresponds to a learning algorithm or to a generic data processing unit. Each node can have a training phase, during which the internal structures are learned from training data (e.g. the weights of a neural network are adapted or the covariance matrix is estimated) and an execution phase, where new data can be processed forwards (by processing the data through the node) or backwards (by applying the inverse of the transformation computed by the node if defined). MDP is designed to make the implementation of new algorithms easy and intuitive, for example by setting automatically input and output dimension and by casting the data to match the typecode (e.g. float or double) of the internal structures. The nodes were designed to be applied to arbitrarily long sets of data: the internal structures can be updated successively by sending chunks of the input data (this is equivalent to online learning if the chunks consists of single observations, or to batch learning if the whole data is sent in a single chunk). Already implemented nodes include Principal Component Analysis (PCA), Independent Component Analysis (ICA), and Slow Feature Analysis (SFA). A flow consists in an acyclic graph of nodes (currently only node sequences are implemented). The data is sent to an input node and is successively processed by the following nodes on the graph. The general flow implementation automatizes the training, execution and inverse execution (if defined) of the whole graph. A subclass of the basic flow class allows user-supplied checkpoint functions to be executed at the end of each phase, for example to save the internal structures of a node for later analysis. Best regards, Pietro Berkes and Tiziano Zito ---------------------------------------- {p.berkes, t.zito}(a)biologie.hu-berlin.de Institute for Theoretical Biology Humboldt University Invalidenstrasse 43 D-10115 Berlin, Germany ----------------------------------------

1 0

ipython, gtk & matplotlib support; end of current ipython development
by Fernando Perez Aug. 22, 2004

Aug. 22, 2004

Hi everyone, sorry for the cross post to those of you who are on all these lists, but since this will affect ipython's future quite a bit, I want a significant heads-up to be seen by all potentially affected. 1. PyGTK & matplotlib --------------------- Thanks to Antoon Pardon and John Hunter, ipython has nearly ready full support for interactive matplotlib with all backends. In this process, we've also added GTK threading support, so you can now use ipython for pygtk development. This code is now in IPython's CVS, and the matplotlib features require matplotlib CVS (for matplotlib use only; matplotlib has NOT become a general ipython requirement). So those of you willing to bleed a little can use it, and now is your opportunity to let us know of any problems you find. Our solution is simpler, but more limited in scope, than scipy's gui_thread. However it currently does NOT work with the WX backends, only with Tk and GTK (-AGG or not). Help from any WX guru is welcome, the place to look is at the end of IPython/Shell.py. I hope that we can find, at least in our more limited context, a working solution for WX which doesn't require all the complexity of gui_thread. In order to use this, do an 'ipython -upgrade' after a cvs update; this will get the necessary support files into ~/.ipython. You will then need to use the new threading options; copying from the docs: -gthread, -mpthread: these are special options, only one of which can be given, and which can ONLY be given as the FIRST option passed to IPython (they will have no effect in any other position). If -gthread is given, IPython starts running a separate thread for GTK operation, so that pyGTK-based programs can open GUIs without blocking IPython. The -mpthread option adds special support for the matplotlib library (http://matplotlib.sourceforge.net), allowing interactive usage of the GTK and WX backends. This also modifies the @run command to correctly execute any matplotlib-based script which calls show() at the end, without blocking. The most convenient way to use this is the new pylab profile, which should be invoked as follows (aliasing this in your shell may be a good idea): $ ipython -mpthread -p pylab pylab will honor your choice of matplotlib backend, though currently it will revert (with a warning) WX to TkAgg, since WX is broken. This will go away once we figure out the WX problems. 2. IPython's future ------------------- Once this support for matplotlib is working to satisfaction, it will mean the end of the line for any more feature-related changes to ipython for quite a while. Once this is reasonably shaken (I hope with at most one more release beyond 0.6.3, which will officially include this), I plan on beginning the long-awaited internal cleanup of ipython. Given my very limited time, this will mean essentially ZERO new features on ipython for quite a while. It will also mean that the new ipython will: - require python 2.3: I want to deprecate as much redundant code as I can from the ipython distribution. I'll use optparse and any other new module from the stdlib which can help shrink ipython. - break backwards compatibility in many areas. In particular, the ipythonrc files will become true python files. - the internal class structure of ipython will drastically change. If you have code which uses ipython via IPython.Shell, you should be fine, as I'll try to keep that API stable. If you've been poking your dirty fingers into iplib or ipmaker directly, expect things to break badly, and don't even think about complaining :) Hopefully once this is over, it will mean having a much cleaner ipython, with an easy path for including it into GUI shells (such as pycrust), and a sane internal code structure. As the new design shakes down, we'll eventually have an ipython 1.0 at last ;) Because of these changes coming down the pipe, if you have any further patches or changes for the current ipython which you'd like to see included, please send them NOW to me. Once I shift gears to the cleanup project, I'll unfortunately have to drop most changes not going in that direction, simply for lack of time. I'd like to thank everybody who has contributed to ipython's development so far, and to encourage others to join in. The cleanup should not be too hard, and it will open the door for having ipython as the interactive core for very high quality python-based environments in the future. Best regards, f

1 0

sparse matrices
by Curzio Basso Aug. 20, 2004

Aug. 20, 2004

Hi all! Can anyone point me to any kind documentation for the scipy.sparse modules? Or I'd better explore it? cheers

2 3

Matplotlib problem under Windows: %HOME% expands to %USERPROFILE%
by Alan G Isaac Aug. 18, 2004

Aug. 18, 2004

I found a Matplotlib problem under Win 2000. THE SYMPTOM: Matplotlib appears to install fine under Windows 2000, but trying to include text in a plot produces the following error message: Traceback (most recent call last): [snip] File "C:\programs\Python23\lib\site-packages\matplotlib\font_manager.py", line 764, in __init__ pickle.dump(self.ttfdict, file(ttfcache, 'w')) IOError: [Errno 2] No such file or directory: '%USERPROFILE%\\.ttffont.cache' THE PROBLEM: In matplotlib's font_manager.py, we find the line ttfpath = os.environ.get('HOME', get_data_path()) (as well as one more problematic use of HOME). Under Windows 2000 (and other Windows?), %HOME% expands to %USERPROFILE%, *not* to the user's home directory. %USERPROFILE% in turn expands to the user's home directory, but this expansion is not done by the line above. COMMENT: this problem affects the installation as well: a directory called literally '%USERPROFILE% is created! THE HACK for Windows users with the problematic install and behavior: change HOME to USERPROFILE in font_manager.py. Hopefully there is also a simple solution for Matplotlib, but I'm just an end user so I won't propose one. Cheers, Alan Isaac

3 4

Install problem
by Wuren Li Aug. 18, 2004

Aug. 18, 2004

Hello- I'm installing SciPy for the first time, on Fedora core 2. I downloaded the rpms, installing (hopefully this is correct, directions aren't clear): Numeric-23.1-1_py23.i386.rpm Scipy_core-0.3.0_108.1820-2_py23_rh9.i386.rpm F2PY-2.39.235_1693-2ee_py22_rh9.noarch.rpm SciPy-0.3.0_266.4239-1ee_py23_Linux_PIII_rh9.i386.rpm The last of these won't install due to dependencies on libfftw.so.2 and librfftw.so.2. Are these necessary, and if so, where is the best place to get them? Thanks! P.S. Numeric-23.1-1_py23.i386.rpm has a small glitch--a dependency on Python 2.3 that prevented installation even though I have 2.3.3. Used --nodeps to force installation. --------------------------------- Do you Yahoo!? Yahoo! Mail - Helps protect you from nasty viruses.

2 1

Matplotlib font encoding problem
by Alan G Isaac Aug. 18, 2004

Aug. 18, 2004

Matplotlib is doing something with its fonts that is unfriendly. E.g., Ghostscript and Distiller choke on http://www.american.edu/econ/problems/matplotlib.fontproblem.eps Ghostscript says it is an invalid front. If I get rid of the verdana font encoding, the file displays fine. I have the same problem when I try to use the TeX fonts. I'm hoping there's a quick fix... Alan Isaac using Matplotlib with Numeric under Python 2.3 OS: Win 2000 Pro

1 0

Re: core dump during import
by Skip Montanaro Aug. 17, 2004

Aug. 17, 2004

A few days ago on scipy-user and earlier today on numpy-discussion I wrote: If I execute from scipy.stats import distribution at the interpreter prompt I get a segfault. Stepping through the suspect code yielded very strange results, so I tried recompiling cephesmodule.c with -O0. That seems to have solved the problem for now. I guess that means it's a GCC optimizer bug. -- Skip Montanaro Got spam? http://www.spambayes.org/ skip(a)pobox.com

1 0

fmin_l_bfgs_b problem
by bulatov＠cs.orst.edu Aug. 16, 2004

Aug. 16, 2004

> On Mon, Aug 09, 2004 at 02:39:59AM -0700, bulatov at cs.orst.edu wrote: > > Was anyone able to use fmin_l_bfgs_b optimizer successfully? > > Well, I wrote the wrapper, so I hope so :-) > > > When I run it with the same parameters as the fmin_bfgs example in the tutorial, > > I get the error: > > > > xopt = fmin_l_bfgs_b(log_likelihood, x0, fprime=log_likelihood_der) > > File "C:\Python23\Lib\site-packages\scipy\optimize\lbfgsb.py", line 174, in fm > > in_l_bfgs_bisave, dsave) > > error: failed in converting 7th argument `g' of _lbfgsb.setulb to C/Fortran > > array > > > Are you sure log_likelihood_der returns an array (of the same length as x0)? I found the problem :) Log_likelihood_der returned Numarray array instead of Numeric array. That happened because I couldn't find "zeros" function in SciPy, so I used numarray.zeros to initialize array. When I changed numarray.zeros to Numeric.zeros, problem was fixed. For the future...how do I debug these kinds of problems? I looked in lbfgs.pyf, but couldn't quite understand why Numeric arrays were required Yaroslav

2 1

ANN: Python/Scientific Job openings at Enthought
by eric jones Aug. 16, 2004

Aug. 16, 2004

Hey Folks, Enthought continues to grow, and we're hiring again. There are several positions open, and most of them involve python+scientific computing. A couple of the positions also involve support of SciPy. I'm very interested in finding a person for the "scientist" job posting that is both able to work well with our customers and also has the personality and interest to work with the SciPy community to help push the library forward. More information is available on our web site. http://www.enthought.com/careers.htm Our current development efforts are in the areas of electromagnetics, geophysics, and graphics (2D and 3D). If you have any questions, please feel free to contact me. Please forward the link to anyone you feel might be interested. thanks, eric

1 0