Mailman 3 November 2008 - 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

scipy/numpy on CentOS 5/ Dependency problem
by Matthias Blaicher 08 Dec '10

08 Dec '10

Hello, I want to install SciPy and numpy on a CentOS 5 system. It contains a fresh install with up-to-date packages. I use the "official" Suse Build repos by DavidCournapeau. It fails with a conflict between refblas3 and blas. [root@localhost yum.repos.d]# yum install python-numpy python-scipy Loading "fastestmirror" plugin Loading mirror speeds from cached hostfile * home_ashigabou: download.opensuse.org * extras: mirrors.tummy.com * updates: dds.gina.alaska.edu * base: mirror.hmc.edu * addons: mirrors.tummy.com extras 100% |=========================| 1.1 kB 00:00 updates 100% |=========================| 951 B 00:00 base 100% |=========================| 1.1 kB 00:00 addons 100% |=========================| 951 B 00:00 Setting up Install Process Parsing package install arguments Resolving Dependencies --> Running transaction check ---> Package python-scipy.i386 0:0.6.0-2.1 set to be updated --> Processing Dependency: libblas.so.3 for package: python-scipy --> Processing Dependency: libblas.so.3 for package: python-scipy --> Processing Dependency: liblapack.so.3 for package: python-scipy --> Processing Dependency: libgfortran.so.1 for package: python-scipy --> Processing Dependency: liblapack.so.3 for package: python-scipy ---> Package python-numpy.i386 0:1.2.0-1.1 set to be updated --> Processing Dependency: gcc-gfortran for package: python-numpy --> Processing Dependency: refblas3 for package: python-numpy --> Processing Dependency: lapack3 < 3.1 for package: python-numpy --> Running transaction check ---> Package gcc-gfortran.i386 0:4.1.2-42.el5 set to be updated --> Processing Dependency: gcc = 4.1.2-42.el5 for package: gcc-gfortran ---> Package blas.i386 0:3.0-37.el5 set to be updated ---> Package lapack3.i386 0:3.0-19.1 set to be updated ---> Package lapack.i386 0:3.0-37.el5 set to be updated ---> Package refblas3.i386 0:3.0-11.1 set to be updated ---> Package libgfortran.i386 0:4.1.2-42.el5 set to be updated --> Running transaction check ---> Package gcc.i386 0:4.1.2-42.el5 set to be updated --> Processing Dependency: libgomp.so.1 for package: gcc --> Processing Dependency: glibc-devel >= 2.2.90-12 for package: gcc --> Processing Dependency: libgomp = 4.1.2-42.el5 for package: gcc --> Running transaction check ---> Package glibc-devel.i386 0:2.5-24 set to be updated --> Processing Dependency: glibc-headers for package: glibc-devel --> Processing Dependency: glibc-headers = 2.5-24 for package: glibc-devel ---> Package libgomp.i386 0:4.1.2-42.el5 set to be updated --> Running transaction check ---> Package glibc-headers.i386 0:2.5-24 set to be updated --> Processing Dependency: kernel-headers for package: glibc-headers --> Processing Dependency: kernel-headers >= 2.2.1 for package: glibc-headers --> Running transaction check ---> Package kernel-headers.i386 0:2.6.18-92.1.13.el5 set to be updated --> Processing Conflict: refblas3 conflicts blas --> Finished Dependency Resolution Error: refblas3 conflicts with blas Is there anything I'm doing wrong here, as I don't want to compile all dependencies by myself.. Sincerly, Matthias Blaicher

3 2

3D plotting
by Robert Cimrman 10 Sep '09

10 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

assign to diagonal values?
by Carlos Scheidegger 07 Jul '09

07 Jul '09

Say I have a matrix m whose diagonal values I want to replace with the contents of vector new_diagonal. Right now, I'm using m -= diag(m) m += diag(new_diagonal) but that seems pretty wasteful (especially if diag() creates new arrays, which I suspect it does). Is there a standard (or any efficient) way to assign the diagonal values directly without an explicit for loop? Googling for "site:www.scipy.org assign diagonal" only returned results for sparse matrices, which is not the case here - these are plain numpy arrays. Thanks in advance, -carlos

11 17

Is numpy's argsort lying about its numpy.int32 types?
by Rob Clewley 15 May '09

15 May '09

Hi, I'm having a problem comparing some types when using numpy's argsort. I'm using numpy 1.0.2. I can reproduce it simply: Python 2.4.4 (#71, Oct 18 2006, 08:34:43) [MSC v.1310 32 bit (Intel)] << snip >> In [1]: from numpy import argsort, sort, int32, array In [2]: x=array([1,3,2]) In [3]: aa=argsort(x) In [4]: as=sort(x) In [5]: type(aa[0]) Out[5]: <type 'numpy.int32'> In [6]: type(as[0]) Out[6]: <type 'numpy.int32'> In [7]: int32 Out[7]: <type 'numpy.int32'> In [8]: type(as[0])==int32 Out[8]: True In [9]: type(aa[0])==int32 Out[9]: False Any of the three indices in aa give me the same problem. Can someone explain if I should be doing this a different way, and if this is a bug?! Thanks, Rob

5 6

signal.lti(A,B,C,D) with D=0
by William Purcell 18 Feb '09

18 Feb '09

I am trying to set up state space representation of a system using signal.lti. The system has no feedforward so D = 0. I've tried the three options listed in the code below to represent D. The only way I can get it to work is option 2 if C has 1 row. If C has more than 1 row it won't work. Any thoughts? -Bill Code --------------------------------------------------------------- from numpy import ones,matrix from scipy import signal r = ones(3) A = matrix([r,r,r]) B = matrix([r]).T C = matrix([r,r]) #three options of D to make it 0 #1) D=0 #2) D = matrix([0,0]).T #3) D = None D = 0 #D = None #D = matrix([0,0]).T Gss = signal.lti(A,B,C,D) ----------------------------------------------------------- Tracebacks ----------------------------------------------------------- Option 1 /usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in abcd_normalize(A, B, C, D) 101 raise ValueError, "A and C must have the same number of columns." 102 if MD != MC: --> 103 raise ValueError, "C and D must have the same number of rows." 104 if ND != NB: 105 raise ValueError, "B and D must have the same number of columns." <type 'exceptions.ValueError'>: C and D must have the same number of rows. WARNING: Failure executing file: <testlti.py> Option 2 (with C as two rows...if C is a single row I do not get this traceback) /usr/lib/python2.5/site-packages/scipy/signal/ltisys.py in ss2tf(A, B, C, D, input) 147 148 num_states = A.shape[0] --> 149 type_test = A[:,0] + B[:,0] + C[0,:] + D 150 num = numpy.zeros((nout, num_states+1), type_test.dtype) 151 for k in range(nout): <type 'exceptions.ValueError'>: shape mismatch: objects cannot be broadcast to a single shape WARNING: Failure executing file: <testlti.py> Option 3 same as 1

4 8

ANN: NumPy 1.2.1
by Jarrod Millman 18 Dec '08

18 Dec '08

I'm pleased to announce the release of NumPy 1.2.1. NumPy is the fundamental package needed for scientific computing with Python. It contains: * a powerful N-dimensional array object * sophisticated (broadcasting) functions * basic linear algebra functions * basic Fourier transforms * sophisticated random number capabilities * tools for integrating Fortran code. Besides it's obvious scientific uses, NumPy can also be used as an efficient multi-dimensional container of generic data. Arbitrary data-types can be defined. This allows NumPy to seamlessly and speedily integrate with a wide-variety of databases. This bugfix release comes almost one month after the 1.2.0 release. Please note that NumPy 1.2.1 requires Python 2.4 or greater. For information, please see the release notes: https://sourceforge.net/project/shownotes.php?release_id=636728&group_id=13… You can download the release from here: https://sourceforge.net/project/showfiles.php?group_id=1369 Thank you to everybody who contributed to this release. Enjoy, -- Jarrod Millman Computational Infrastructure for Research Labs 10 Giannini Hall, UC Berkeley phone: 510.643.4014 http://cirl.berkeley.edu/

3 11

scikits.timeseries
by Robert Ferrell 11 Dec '08

11 Dec '08

Timeseries is an awesome package. Great contribution. I have 2 questions about it, though. 1. Is scipy-user the right place for questions? 2. I've noticed that 'business frequency' includes holidays, and that can create holes in what are actually complete data sets. For instance, Sep 01, 2008 was a holiday in the US (Labor Day). However, it is included in a DateArray spanning that date. For instance. In [640]: ts.date_array(ts.Date('B','2008-08-25'), length=12) Out[640]: DateArray([25-Aug-2008, 26-Aug-2008, 27-Aug-2008, 28-Aug-2008, 29- Aug-2008, 01-Sep-2008, 02-Sep-2008, 03-Sep-2008, 04-Sep-2008, 05-Sep-2008, 08-Sep-2008, 09-Sep-2008], freq='B') This makes stock ticker data look like it's incomplete - no data for Sep 01, since the markets were closed. For instance, if I use matplotlib.finance.quotes_historical_yahoo to download Intel data, and put that into the date array above, I get the series: masked_array(data = [22.77 22.95 23.21 23.39 22.67 -- 22.39 21.35 20.34 20.43 20.79], mask = [False False False False False True False False False False False], fill_value=1e+20) That has a hole on Sep 1. This matters for things like moving average calculation. Sep 1 should be treated like a Saturday or Sunday, but instead causes a 5-day mov_average calculation to not compute anything from Sep 2 through Sep 7: timeseries([-- -- -- -- 22.998 -- -- -- -- -- 21.06], dates = [25-Aug-2008 ... 08-Sep-2008], freq = B) My question: What is a good way to handle (get rid of?) the holes in the series? thanks, -robert

3 10

new Kolmogorov-Smirnov test
by josef.pktd＠gmail.com 03 Dec '08

03 Dec '08

Since the old scipy.stats.kstest wasn't correct, I spent quite some time fixing and testing it. Now, I know more about the Kolmogorov-Smirnov test, than I wanted to. The kstest now resembles the one in R and in matlab, giving the option for two-sided or one-sided tests. The names of the keyword options are a mixture of matlab and R, which I liked best. Since the exact distribution of the two-sided test is not available in scipy, I use an approximation, that seems to work very well. In several Monte Carlo studies against R, I get very close results, especially for small p-values. (For those interested, for small p-values, I use ksone.sf(D,n)*2; for large p-values or large n, I use the asymptotic distribution kstwobign) example signature and options: kstest(x,testdistfn.name, alternative = 'unequal', mode='approx')) kstest(x,testdistfn.name, alternative = 'unequal', mode='asymp')) kstest(x,testdistfn.name, alternative = 'larger')) kstest(x,testdistfn.name, alternative = 'smaller')) Below is the Monte Carlo for the case, when the random variable and the hypothesized distribution both are standard normal (with sample size 100 and 1000 replications. Rejection rates are very close to alpha levels. It also contains the mean absolute error MAE for the old kstest. I also checked for mean shifted normal random variables. In all cases that I tried, I get exactly the same rejection rates as in R. For details see doc string or source. I attach file to a separate email, to get around attachment size limit. I intend to put this in trunk tomorrow, review and comments are welcome. Josef data generation distribution is norm, hypothesis is norm ================================================== n = 100, loc = 0.000000 scale = 1.000000, n_repl = 1000 columns: D, pval rows are kstest(x,testdistfn.name, alternative = 'unequal', mode='approx')) kstest(x,testdistfn.name, alternative = 'unequal', mode='asymp')) kstest(x,testdistfn.name, alternative = 'larger')) kstest(x,testdistfn.name, alternative = 'smaller')) Results for comparison with R: MAE old kstest [[ 0.00453195 0.19152727] [ 0.00453195 0.2101139 ] [ 0.02002774 0.19145982] [ 0.02880553 0.26650226]] MAE new kstest [[ 1.87488913e-17 1.07738517e-02] [ 1.87488913e-17 1.91763848e-06] [ 2.38576520e-17 8.90287843e-16] [ 1.41312743e-17 9.92428362e-16]] percent count absdev > 0.005 [[ 0. 53.9] [ 0. 0. ] [ 0. 0. ] [ 0. 0. ]] percent count absdev > 0.01 [[ 0. 24.3] [ 0. 0. ] [ 0. 0. ] [ 0. 0. ]] percent count abs percent dev > 1% [[ 0. 51.8] [ 0. 0. ] [ 0. 0. ] [ 0. 0. ]] percent count abs percent dev > 10% [[ 0. 0.] [ 0. 0.] [ 0. 0.] [ 0. 0.]] new: count rejection at 1% significance [ 0.01 0.008 0.009 0.014] R: proportion of rejection at 1% significance [ 0.01 0.008 0.009 0.014] new: proportion of rejection at 5% significance [ 0.054 0.048 0.048 0.06 ] R: proportion of rejection at 5% significance [ 0.054 0.048 0.048 0.06 ] new: proportion of rejection at 10% significance [ 0.108 0.096 0.095 0.109] R: proportion of rejection at 10% significance [ 0.108 0.096 0.095 0.109]

4 8

os x, intel compilers & mkl, and fink python
by Gideon Simpson 02 Dec '08

02 Dec '08

Has anyone gotten the combination of OS X with a fink python distribution to successfully build numpy/scipy with the intel compilers and the mkl? If so, how'd you do it? -gideon

3 4