Mailman 3 January 2011 - NumPy-Discussion

bincount question
by Alan G Isaac 09 Jan '11

09 Jan '11

:: >>> np.bincount([]) Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: The first argument cannot be empty. Why not? (I.e., why isn't an empty array the right answer?) Thanks, Alan Isaac

2 1

speed of numpy.ndarray compared to Numeric.array
by EMMEL Thomas 08 Jan '11

08 Jan '11

Hi, There are some discussions on the speed of numpy compared to Numeric in this list, however I have a topic I don't understand in detail, maybe someone can enlighten me... I use python 2.6 on a SuSE installation and test this: #Python 2.6 (r26:66714, Mar 30 2010, 00:29:28) #[GCC 4.3.2 [gcc-4_3-branch revision 141291]] on linux2 #Type "help", "copyright", "credits" or "license" for more information. import timeit #creation of arrays and tuples (timeit number=1000000 by default) timeit.Timer('a((1.,2.,3.))','from numpy import array as a').timeit() #8.2061841487884521 timeit.Timer('a((1.,2.,3.))','from Numeric import array as a').timeit() #9.6958281993865967 timeit.Timer('a((1.,2.,3.))','a=tuple').timeit() #0.13814711570739746 #Result: tuples - of course - are much faster than arrays and numpy is a bit faster in creating arrays than Numeric #working with arrays timeit.Timer('d=x1-x2;sum(d*d)','from Numeric import array as a; x1=a((1.,2.,3.));x2=a((2.,4.,6.))').timeit() #3.263314962387085 timeit.Timer('d=x1-x2;sum(d*d)','from numpy import array as a; x1=a((1.,2.,3.));x2=a((2.,4.,6.))').timeit() #9.7236979007720947 #Result: Numeric is three times faster than numpy! Why? #working with components: timeit.Timer('d0=x1[0]-x2[0];d1=x1[1]-x2[1];d2=x1[2]-x2[2];d0*d0+d1*d1+d2*d2','a=tuple; x1=a((1.,2.,3.));x2=a((2.,4.,6.))').timeit() #0.64785194396972656 timeit.Timer('d0=x1[0]-x2[0];d1=x1[1]-x2[1];d2=x1[2]-x2[2];d0*d0+d1*d1+d2*d2','from numpy import array as a; x1=a((1.,2.,3.));x2=a((2.,4.,6.))').timeit() #3.4181499481201172 timeit.Timer('d0=x1[0]-x2[0];d1=x1[1]-x2[1];d2=x1[2]-x2[2];d0*d0+d1*d1+d2*d2','from Numeric import array as a; x1=a((1.,2.,3.));x2=a((2.,4.,6.))').timeit() #0.97426199913024902 Result: tuples are again the fastest variant, Numeric is faster than numpy and both are faster than the variant above using the high-level functions! Why? For various reasons I need to use numpy in the future where I used Numeric before. Is there any better solution in numpy I missed? Kind regards and thanks in advance Thomas This email and any attachments are intended solely for the use of the individual or entity to whom it is addressed and may be confidential and/or privileged. If you are not one of the named recipients or have received this email in error, (i) you should not read, disclose, or copy it, (ii) please notify sender of your receipt by reply email and delete this email and all attachments, (iii) Dassault Systemes does not accept or assume any liability or responsibility for any use of or reliance on this email.For other languages, go to http://www.3ds.com/terms/email-disclaimer.

3 2

code for multidimensional scaling?
by Brian Murphy 07 Jan '11

07 Jan '11

Hi, I'm new to the list, so I hope my question is appropriate (I've already sent the same posting to the SciPy Users list). I'm looking for code that implements multi-dimensional scaling (e.g. like Matlab's mdscale command) in Python. My best guess was that I would find it in the Scikit Learn package, but couldn't turn anything up. Any suggestions? thanks and regards, Brian -- Brian Murphy Post-Doctoral Researcher Language, Interaction and Computation Lab Centre for Mind/Brain Sciences University of Trento http://clic.cimec.unitn.it/brian/

1 0

aa.astype(int) truncates and doesn't round
by josef.pktd＠gmail.com 07 Jan '11

07 Jan '11

just something I bumped into and wasn't aware of >>> aa array([ 1., 1., 1., 1., 1.]) >>> aa.astype(int) array([0, 1, 0, 0, 0]) >>> aa - 1 array([ -2.22044605e-16, 2.22044605e-16, -2.22044605e-16, -3.33066907e-16, -3.33066907e-16]) >>> np.round(aa).astype(int) array([1, 1, 1, 1, 1]) Josef

3 4

NEP for faster ufuncs
by Mark Wiebe 06 Jan '11

06 Jan '11

Hello NumPy-ers, After some performance analysis, I've designed and implemented a new iterator designed to speed up ufuncs and allow for easier multi-dimensional iteration. The new code is fairly large, but works quite well already. If some people could read the NEP and give some feedback, that would be great! Here's a link: https://github.com/m-paradox/numpy/blob/mw_neps/doc/neps/new-iterator-ufunc… I would also love it if someone could try building the code and play around with it a bit. The github branch is here: https://github.com/m-paradox/numpy/tree/new_iterator To give a taste of the iterator's functionality, below is an example from the NEP for how to implement a "Lambda UFunc." With just a few lines of code, it's possible to replicate something similar to the numexpr library (numexpr still gets a bigger speedup, though). In the example expression I chose, execution time went from 138ms to 61ms. Hopefully this is a good Christmas present for NumPy. :) Cheers, Mark Here is the definition of the ``luf`` function.:: def luf(lamdaexpr, *args, **kwargs): """Lambda UFunc e.g. c = luf(lambda i,j:i+j, a, b, order='K', casting='safe', buffersize=8192) c = np.empty(...) luf(lambda i,j:i+j, a, b, out=c, order='K', casting='safe', buffersize=8192) """ nargs = len(args) op = args + (kwargs.get('out',None),) it = np.newiter(op, ['buffered','no_inner_iteration'], [['readonly','nbo_aligned']]*nargs + [['writeonly','allocate','no_broadcast']], order=kwargs.get('order','K'), casting=kwargs.get('casting','safe'), buffersize=kwargs.get('buffersize',0)) while not it.finished: it[-1] = lamdaexpr(*it[:-1]) it.iternext() return it.operands[-1] Then, by using ``luf`` instead of straight Python expressions, we can gain some performance from better cache behavior.:: In [2]: a = np.random.random((50,50,50,10)) In [3]: b = np.random.random((50,50,1,10)) In [4]: c = np.random.random((50,50,50,1)) In [5]: timeit 3*a+b-(a/c) 1 loops, best of 3: 138 ms per loop In [6]: timeit luf(lambda a,b,c:3*a+b-(a/c), a, b, c) 10 loops, best of 3: 60.9 ms per loop In [7]: np.all(3*a+b-(a/c) == luf(lambda a,b,c:3*a+b-(a/c), a, b, c)) Out[7]: True

7 27

newbie question (curve fitting Z=f(X,Y))
by Sachin Kumar Sharma 06 Jan '11

06 Jan '11

Hi, Absolute basic question I want to do following * Read data from excel sheet (three columns - X,Y,Z) * Curve fit a function Z=F(X,Y) * Plot X Vs Z (from data) and plot X Vs Z (from curve fit) Kindly advise me how to write a basic python script for the same. Thanks & Regards Sachin ************************************************************************ Sachin Kumar Sharma Senior Geomodeler

2 1

Question regarding submitting patches
by Justin Peel 06 Jan '11

06 Jan '11

Hi all, I've been submitting some patches recently just by putting them on Trac. However, I noticed in the Numpy Developer Guide that it says: The recommended way to proceed is either to attach these files to an enhancement ticket in the Numpy Trac and send a mail about the enhancement to the NumPy mailing list. This line is rather confusing. Either the 'either' should be removed or 'and'->'or'. In other words, is it sufficient to submit the patch in Trac or should I also be emailing the Numpy mailing list about each patch I submit? Justin

4 3

NumPy on HPC podcast
by Brock Palen 05 Jan '11

05 Jan '11

I host and HPC podcast with Jeff Squyres of OpenMPI fame: www.rce-cast.com We would like to have a developer or two from NumPy on the show to represent the project. We do this over phone or skype and takes about an hour. Feel free to contact me out of band. I hope to hear from you soon! Brock Palen www.umich.edu/~brockp Center for Advanced Computing brockp(a)umich.edu (734)936-1985

2 2

Giving numpy the ability to multi-iterate excluding an axis
by John Salvatier 05 Jan '11

05 Jan '11

A while ago, I asked a whether it was possible to multi-iterate over several ndarrays but exclude a certain axis( http://www.mail-archive.com/numpy-discussion@scipy.org/msg29204.html), sort of a combination of PyArray_IterAllButAxis and PyArray_MultiIterNew. My goal was to allow creation of relatively complex ufuncs that can allow reduction or directionally dependent computation and still use broadcasting (for example a moving averaging ufunc that can have changing averaging parameters). I didn't get any solutions, which I take to mean that no one knew how to do this. I am thinking about trying to make a numpy patch with this functionality, and I have some questions: 1) How difficult would this kind of task be for someone with non-expert C knowledge and good numpy knowledge? 2) Does anyone have advice on how to do this kind of thing? Best Regards, John

2 9

Submitting patches
by Justin Peel 04 Jan '11

04 Jan '11

Hi all, I've been submitting some patches recently just by putting them on Trac. However, I noticed in the Numpy Developer Guide that it says: The recommended way to proceed is either to attach these files to an enhancement ticket in the Numpy Trac and send a mail about the enhancement to the NumPy mailing list. This line is rather confusing. Either the 'either' should be removed or 'and'->'or'. In other words, is it sufficient to submit the patch in Trac or should I also be emailing the Numpy mailing list about each patch I submit? Justin

1 0