Mailman 3 February 2006 - SciPy-User

Help required regding Eigenvectors
by Sudheer Phani 14 Mar '22

14 Mar '22

Hello All, I have been encountering some strange problems with the calculation of eigen vectors using python. The problem is with the sign of the eigen vectors calculated. On Very few occasions the eigen vectors calculated using python have an opposite sign than the actual eigenvectors, though the magnitude is correct. Can any one of you plz help me in this regard and suggest me some solution if you have encoutered similar problems . Any help in this regd is appreciated Thanks in Advace regds Sudheer

5 5

3D plotting
by Robert Cimrman 11 Sep '09

11 Sep '09

Hi, I need to do some simple 3D plots, so I tried the two Cookbook recipes (xplt, mplot3d) and failed: * xplt: does not work when using a custom installation directory such as python setup.py install --root=/home/share/software apparently the gistdata directory path is wrong: gist.error: failed to create drawing -- Gist work.gs style sheet missing * mplot3d: does not work with my version of matplotlib (0.80). I have made the changes mentioned in the Cookbook to no avail. (Axes.__init__() args apparently changed, as well as some other matplotlib object attributes) Any ideas? mplot3d looks great, I would really like to use it! thanks, r.

6 6

scipy.test problem
by Hanno Klemm 17 Mar '06

17 Mar '06

Hi there, I am new to scipy and have the following installation problem: I work on a dual processor Xeon box, and try to install scipy. After various unsuccessful attempts, I now compiled BLAS, LAPACK ATLAS and FFTW3.1 from scratch, then installed numpy-0.9.5 and then installed scipy-0.4.6 under python 2.4.2. The installation went fine but after import scipy scipy.test(level=1) I get the following error message: ====================================================================== FAIL: check_simple (scipy.linalg.decomp.test_decomp.test_eig) ---------------------------------------------------------------------- Traceback (most recent call last): File "/scratch/python2.4/lib/python2.4/site-packages/scipy/linalg/tests/test_decomp.py", line 89, in check_simple assert_array_almost_equal(w,exact_w) File "/scratch/python2.4/lib/python2.4/site-packages/numpy/testing/utils.py", line 183, in assert_array_almost_equal assert cond,\ AssertionError: Arrays are not almost equal (mismatch 66.6666666667%): Array 1: [ 9.4371906+0.j -0.1153653+0.j -0.3218254+0.j] Array 2: [ 9.3218254 0. -0.3218254] ====================================================================== FAIL: check_simple (scipy.linalg.decomp.test_decomp.test_eigvals) ---------------------------------------------------------------------- Traceback (most recent call last): File "/scratch/python2.4/lib/python2.4/site-packages/scipy/linalg/tests/test_decomp.py", line 37, in check_simple assert_array_almost_equal(w,exact_w) File "/scratch/python2.4/lib/python2.4/site-packages/numpy/testing/utils.py", line 183, in assert_array_almost_equal assert cond,\ AssertionError: Arrays are not almost equal (mismatch 66.6666666667%): Array 1: [ 9.4371906+0.j -0.1153653+0.j -0.3218254+0.j] Array 2: [ 9.3218254 0. -0.3218254] ====================================================================== FAIL: check_simple_tr (scipy.linalg.decomp.test_decomp.test_eigvals) ---------------------------------------------------------------------- Traceback (most recent call last): File "/scratch/python2.4/lib/python2.4/site-packages/scipy/linalg/tests/test_decomp.py", line 45, in check_simple_tr assert_array_almost_equal(w,exact_w) File "/scratch/python2.4/lib/python2.4/site-packages/numpy/testing/utils.py", line 183, in assert_array_almost_equal assert cond,\ AssertionError: Arrays are not almost equal (mismatch 66.6666666667%): Array 1: [ 9.4371906+0.j -0.1153653+0.j -0.3218254+0.j] Array 2: [ 9.3218254 0. -0.3218254] ---------------------------------------------------------------------- Ran 1067 tests in 2.250s FAILED (failures=3) <unittest.TextTestRunner object at 0x2aa03923d0> Could anybody please tell me what to do with that? If that is important: I compiled everything with gcc. gcc --version gives: gcc (GCC) 3.2.3 20030502 (Red Hat Linux 3.2.3-52) In order to get scipy to use the atlas libraries I had to add an option -fPIC in the compiler flags in atlas and friends. Any help would be much appreciated! Thanks in advance, Hanno -- Hanno Klemm klemm(a)phys.ethz.ch

5 9

New sparse matrix functionality
by Ed Schofield 02 Mar '06

02 Mar '06

Hi all, I'd created a new SVN branch with some new functionality and changes to sparse matrices. This post describes what's new and what's different versus the main branch. If you have an existing SVN tree, you can pull in my branch with the command "svn switch http://svn.scipy.org/svn/scipy/branches/ejs". One of the conclusions of Jonathan Guyer's comparison between scipy.sparse and PySparse in November (http://article.gmane.org/gmane.comp.python.scientific.devel/3187/match=scip… and http://old.scipy.org/wikis/featurerequests/SparseSolvers) was that SciPy's support for efficient construction of sparse matrices is weak. My patch adds a new data type, lil_matrix, that stores non-zeros as a list of sorted Python lists. For the simple benchmark of creating a new 10^3 x 10^5 matrix with 10^4 non-zero elements in random locations and converting to CSR and CSC matrices, the lil_matrix format is slightly more than twice as fast as dok_matrix. It's a row-wise format, so conversion to CSR is very fast, whereas conversion to CSC goes through CSR internally. Index lookups use binary searches, so they take log time. I think the implementation is already complete enough for most uses -- so please try it out and tell me how you go! Another of Jonathan's observations was that SciPy had no support for slicing matrices with A[i, :] like in PySparse. My patch adds support for slice notation and NumPy-style fancy indexing to dok_matrix and lil_matrix objects. With this it's possible to build sparse matrices quickly -- for example: >>> a = lil_matrix((3,5)) >>> a[1,[2,3,4]] = range(3) >>> a[2,:] = 3 * a[1,:] A third new feature is a .sum() method, which takes a single axis argument like in NumPy. My branch also changes one aspect of the existing behaviour: the todense() method now returns a dense (NumPy) matrix, rather than a dense array. Converting to a dense array is now available under a toarray() method (or .A attribute) instead. The rationale behind this change is to emphasize that sparse matrices are closer to dense matrices than to dense arrays in their attributes (e.g. .T for transpose) and behaviour (e.g. * means inner product). I've also been careful to make multiplication between sparse matrices and dense row or column vectors (matrix objects with unit length or height) return matrices of the correct dimensions, rather than arrays. Several unit tests relied on the old behaviour, and I've changed these accordingly. Most of these test changes are just simplifications -- for example assert_array_equal((a*c).todense(), a.todense() * c) instead of assert_array_equal((a*c).todense(), dot(a.todense(), c)) -- but I'd appreciate some criticism and feedback on which behaviour people prefer. These changes have highlighted a problem present in both the main trunk and my branch: that multiplying a dense matrix 'a' by a sparse matrix 'b' is not possible using the syntax 'a*b'. I'll follow this up with a proposal to numpy-discussion on how we can solve this. -- Ed

3 6

ATLAS and gcc v4.0.2
by Nils Wagner 01 Mar '06

01 Mar '06

Hi all, Where can I find some hints w.r.t. to ATLAS and scipy/numpy ? AFAIK the new website has no information on building ATLAS. Do I need version 3 of gcc ? I am on SuSE Linux 10.0 with gcc v4.0.2 If I try to install ATLAS I get .. Looking for compilers (this may take a while): /usr/bin/gcc : v4.0.2 /usr/bin/gcc : v4.0.2 F77 = /usr/bin/g77 -fomit-frame-pointer -O -m64 CC = /usr/bin/gcc -fomit-frame-pointer -O -mfpmath=387 -m64 MCC = /usr/bin/gcc -fomit-frame-pointer -O -mfpmath=387 -m64 ATLAS has detected that you are running on an x86 architecture (eg., Athlon, P4, PIII, etc), but do not have a pre-2.96 gcc installed. Using gcc 3/2.96 will result in a performance loss of 10-20% IN THE BEST CASE over a gcc previous to 2.96. If your install goes slightly wrong, performance drops of 50% are more normal with these compilers. Full details of this problem can be found at: http://www.cs.utk.edu/~rwhaley/ATLAS/gcc30.html In the meantime, we suggest you abondon this ATLAS install, and install a gcc that delivers decent x86 performance, and restart the install then. You do not need to be root to install, and full instructions, along with all needed files, are given in the ATLAS errata file at: http://math-atlas.sourceforge.net/errata.html#UnixGccInst Stop ATLAS install? [y]: How can I continue ? Nils

6 15

Re: [SciPy-user] ATLAS and gcc v4.0.2
by Eric Emsellem 28 Feb '06

28 Feb '06

Hi, to summarize: just go on with gcc4. It works (although things may not be optimal, but I would guess the difference may be only 10% or so). ==> below is the way I compile things on Suse 10 with gcc4 starting with Lapack and going up to matplotlib. I have no problem with it although compiling ATLAS is a real pain (it just takes forever). Hope this helps. Please adapt things to your machine and system (using the right architecture name for ATLAS). Things that I don't do anymore are commands starting with "#". ALso please update some of the version numbers (ipython, matplotlib). A cleaned and updated version of such a simple procedure would be very helpful, if someone can post it. cheers Eric ============================= ####################### # LAPACK # ##################### # Build Lapack from the web since then you are sure it is complete # It is in the Suse distribution but I don't trust it wget http://www.netlib.org/lapack/lapack.tgz mdkir install_lapack cd install_lapack tar xvfz ../lapack.tgz cp INSTALL/make.inc.LINUX make.inc ## This will build a lapack_LINUX.a ## That you will use later on for ATLAS make lapacklib ######################## # ATLAS # ###################### tar xvfz atlas3.6.0.tar.gz cd ATLAS ## You will have to answer a few questions... # don't care about the warning on gcc make # change "arch" with the right one # you will be asked a few question there # don't care about the warning on gcc make install arch=Linux_UNKNOWNSSE2 ## Create a new directory where to put the include and lib files ## (root) mkdir /usr/local/lib/atlas ## then copy all the include and archive (*.a) files ## under this new dir cp include/* /usr/local/lib/atlas cp *.a /usr/local/lib/atlas # in .cshrc # and don't forget to set it up, this is for tcsh. For bash use "set .. = .." setenv ATLAS /usr/local/lib/atlas ## save the lapack lib built from ATLAS (which is incomplete) cd /usr/local/lib/atlas cp liblapack.a liblapack.a_ATLAS # then create a new temporary dir which will allow you to complete # lapack mkdir tmp cd tmp # detar the atlas lapack lib ar x ../liblapack.a # then cp the true lapack linux lib into the liblapack.a file cp <path>/install_lapack/lapack_LINUX.a ../liblapack.a # and add the files from the atlas lib # you will then have a full lapack lib with the atlas files too ar r ../liblapack.a *.o cd .. # remove the temp dir rm -rf tmp ## copy all *.a into your local atlas dir cp *.a /usr/local/lib/atlas # ###################### ## Added all fftw libraries from yast2 # ###################### ############################################################## ###### NO NEED TO INSTALL NUMERIC IF YOU HAVE ALREADY the new NUMPY... ##### HOWEVER Numarray is required for pyfits ##### NUmeric ###cvs -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/numpy login ###cvs -z3 -d:pserver:anonymous@cvs.sourceforge.net:/cvsroot/numpy co -P Numerical ### #### THEN EDIT your customize.py ############################################################## ##### Customization variables for installing Numeric ### ###use_system_lapack = 1 ###lapack_library_dirs = ['/usr/local/lib/atlas'] ###lapack_libraries = ['lapack', 'cblas', 'f77blas', 'atlas', 'g2c'] ###lapack_extra_link_args = [] ### #### Set use_dotblas to 1 to use BLAS for the matrix multiplication routines. #### Put the directory that 'cblas.h' is in into dotblas_include_dirs ###use_dotblas = 1 ###dotblas_include_dirs = ['/usr/local/lib/atlas'] ###dotblas_cblas_header = '<cblas.h>' ###dotblas_library_dirs = lapack_library_dirs ###dotblas_libraries = lapack_libraries ###dotblas_extra_link_args = [] ############################################################## ### ###python setup.py install ########################################################### # NUMPY # ######### svn co http://svn.scipy.org/svn/numpy/trunk numpy cd numpy python setup.py install # ###################### ## INSTALLED IPYTHON # ###################### tar xvfz ipython-0.6.15.tar.gz cd ipython-0.6.15 python setup.py install # ###################### ## F2PY # ###################### (CVS password: guest) cvs -d :pserver:anonymous@cens.ioc.ee:/home/cvs login cvs -z6 -d :pserver:anonymous@cens.ioc.ee:/home/cvs checkout f2py2e cd f2py2e make install # ###################### ## SCIPY # ###################### ## Remove any scipy in your python before.. ## ALSO REMOVE the "build" directories from the "core" and "scipy" directories before ## doing the install!! svn co http://svn.scipy.org/svn/scipy/trunk scipy cd scipy python setup.py install ## Test:: cd # Start python and run as tests: # (level 10 will take a while ...) import scipy scipy.test(1,verbosity=10) scipy.test(10,verbosity=10) ################################# # Matplotlib 0.86 ################################# # Note that some matplotlibrc options have changed # such that tick ==> xtick + ytick # add one line in setupext.py: o.tcl_inc="/usr/include/" # before if not os.path.exists(o.tcl_inc): print 'cannot find tcl/tk headers. giving up.' sys.exit() return o # and run the usual python setup.py install -- =============================================================== Observatoire de Lyon emsellem(a)obs.univ-lyon1.fr 9 av. Charles-Andre tel: +33 4 78 86 83 84 69561 Saint-Genis Laval Cedex fax: +33 4 78 86 83 86 France http://www-obs.univ-lyon1.fr/eric.emsellem ===============================================================

1 0

poly1d equality test
by Ryan Krauss 28 Feb '06

28 Feb '06

Is there an easy way to test for the equality of two poly1d objects? I have the following problem: In [36]: p1 Out[36]: poly1d([ 6.66666667, 0. , -49. ]) In [37]: p2 Out[37]: poly1d([ 6.66666667, 0. , -49. ]) In [38]: p1==p2 Out[38]: False In [39]: (p1.coeffs==p2.coeffs).all() Out[39]: True It seems like == is only testing if they are the same object. I am trying to write a class for transfer functions which model a system as a numerator and denominator poly1d. When I multiply or divide them I need to check if the numerator and denominators cancel exactly. Thanks, Ryan

3 2

Re: [SciPy-user] [Matplotlib-users] switch to subversion tomorrow Tuesday 28 Feb
by Vincent Favre-Nicolin 28 Feb '06

28 Feb '06

Hi, > I will endeavor to switch matplotlib to a subversion repository (newly > available courtesy of SourceForge) tomorrow afternoon/evening > (US/Pacific time) unless I hear pleas from developers with unmerged > changes. > > I believe the new command for anonymous checkout of matplotlib will be: > svn co https://svn.sourceforge.net/svnroot/matplotlib/trunk/matplotlib > > I'll post the developer checkout command on the matplotlib-devel list > when I figure it out. It's actually the same command. Only, when committing changes you have to supply username and password using, e.g. --username and --password options. But a subversion GUI (I'm happy with esvn for linux & windows) will do that for you. What's really nice, there are no more delay between developers and anonymous access. Vincent -- Vincent Favre-Nicolin Université Joseph Fourier http://v.favrenicolin.free.fr ObjCryst & Fox : http://objcryst.sourceforge.net

2 1

Re: [SciPy-user] Confused about enhanced-scalar dtypes
by LATORNELL, Doug 28 Feb '06

28 Feb '06

Thanks, Travis. I'm not sure how I got it into my head that the dtype argument took the name of the type rather than the type object. The latter is certainly obvious to me now when I go back and read the Data-Type Descriptors section in the NumPy book. dtype=float it shall be! Doug > -----Original Message----- > From: scipy-user-bounces(a)scipy.net > [mailto:scipy-user-bounces@scipy.net] On Behalf Of Travis Oliphant > Sent: February 24, 2006 16:17 > To: SciPy Users List > Subject: Re: [SciPy-user] Confused about enhanced-scalar dtypes > > LATORNELL, Doug wrote: > > >I built NumPy from SVN this morning and got a surprise > concerning dtype > >arguments: > > > >In [24]: import numpy > > > >In [25]: numpy.__version__ > >Out[25]: '0.9.6.2165' > > > >In [26]: a = numpy.fromfile("profiles-O0", dtype="float", sep=' ') > >------------------------------------------------------------- > ---------- > >- > >--- > >exceptions.TypeError Traceback (most > >recent call last) > > > >/ocean/dlatorne/SoG/dlatornell06/profiles/<ipython console> > > > >TypeError: data type not understood > > > > > > I don't recommend using strings like "float". Just use float > (which is a type-object). > > I don't think "float" is actually an accepted string. There > are a few strings that are recognized and more could be > added, but I would just use dtype=float. > > >Based on the tip in the middle of pg 19 of my copy of the > NumPy book, I > >expected both dtype="float" and dtype="float_" to work. > > > > > dtype=float and dtype=numpy.float_ are similar. Note the > absence of "" > > >I know there has been a bunch of work lately on things related to > >dtype, but I don't recall seeing anything about breaking the > >equivalence of the enhanced-scalar types (like float_) and the > >corresponding standard Python types (float, in this case). > Did I miss, > >or misunderstand something, get unluck with the timing of my > checkout > >this morning, or is this a bug? > > > > > I'm surprised "float" worked before and then doesn't work > now. But, there have been enough changes between 0.9.3 and > 0.9.5 that it's > possible. But, don't do it that way :-) > > -Travis > > _______________________________________________ > SciPy-user mailing list > SciPy-user(a)scipy.net > http://www.scipy.net/mailman/listinfo/scipy-user > This email and any files transmitted with it may contain privileged or confidential information and may be read or used only by the intended recipient. If you are not the intended recipient of the email or any of its attachments, please be advised that you have received this email in error and any use, dissemination, distribution, forwarding, printing or copying of this email or any attached files is strictly prohibited. If you have received this email in error, please immediately purge it and all attachments and notify the sender by reply email or contact the sender at the number listed.

1 0

Missing 'plt'
by Alec Edgington 27 Feb '06

27 Feb '06

Hi, I'm hoping someone can help me get started with plotting using 'plt' in scipy. Aplogogies in advance for being an idiot. I'm running on up-to-date Windows XP; I've downloaded and installed python-2.4.2, numpy-0.9.5, wxPython-2.6 and scipy-0.4.6. But the 'plt' sub-package just doesn't seem to be there. Nor does gui_thread. So when I start up IDLE and try to follow the instructions in http://sg0.chem.upenn.edu/adam/howto/scipy/mini-plt.shtml, all that happens is: >>> import gui_thread Traceback (most recent call last): File "<pyshell#0>", line 1, in -toplevel- import gui_thread ImportError: No module named gui_thread >>> from scipy import * >>> from scipy import plt Traceback (most recent call last): File "<pyshell#2>", line 1, in -toplevel- from scipy import plt ImportError: cannot import name plt >>> I checked in C:\Python24\Lib\site-packages\scipy and found no folder called 'plt'. What am I doing wrong? Many thanks, Alec

3 3