Mailman 3 April 2009 - NumPy-Discussion

Replacing colours in numpy array 2
by Frank Peacock 10 Apr '09

10 Apr '09

Hello Apologies for the incorrect posting before. The problem was that I was not using an array for the true condition and instead was using a scalar. Frank

1 0

LARS algorithm
by Gael Varoquaux 10 Apr '09

10 Apr '09

Has anybody implemented the LARS[1] (least angle regression) algorithm for regularized regression in Python or in C binded in Python? I am about to start in such endeaviour, but I wanted to check if someone was willing to share code under a BSD-compatible license. Gaël [1] http://en.wikipedia.org/wiki/Least-angle_regression PS: Sorry for the cross-post, I just wanted to be really sure that if anybody was willing to share such code, I would reach him :)

2 1

Replacing colours in numpy array
by Frank Peacock 10 Apr '09

10 Apr '09

Hello I have a numpy array that I obtained from a converted RGBA gif image. I have tried to replace some colours with different ones using the where condition but have a problem with dimensions. If I use b=where(a==0,255,a) where a is a numpy array from an image it does replace components of the RGB values in each pixel but it fails if I try b=where(a==(0,0,0,255),(255,255,255,255),a) with incorrect dimension error. Could you please help? Thanks Frank -----Original Message----- From: numpy-discussion-bounces(a)scipy.org [mailto:numpy-discussion-bounces@scipy.org] On Behalf Of numpy-discussion-request(a)scipy.org Sent: 10 April 2009 08:18 To: numpy-discussion(a)scipy.org Subject: Numpy-discussion Digest, Vol 31, Issue 24 Send Numpy-discussion mailing list submissions to numpy-discussion(a)scipy.org To subscribe or unsubscribe via the World Wide Web, visit http://mail.scipy.org/mailman/listinfo/numpy-discussion or, via email, send a message with subject or body 'help' to numpy-discussion-request(a)scipy.org You can reach the person managing the list at numpy-discussion-owner(a)scipy.org When replying, please edit your Subject line so it is more specific than "Re: Contents of Numpy-discussion digest..." Today's Topics: 1. Re: DVCS at PyCon (Ondrej Certik) 2. Re: using reducing functions without eliminating dimensions? (Dan Lenski) 3. Re: Another Array (Ian Mallett) 4. Re: Another Array (Robert Kern) 5. Re: Another Array (Ian Mallett) 6. Re: Another Array (Robert Kern) 7. Re: Another Array (Ian Mallett) 8. Re: Another Array (Anne Archibald) ---------------------------------------------------------------------- Message: 1 Date: Thu, 9 Apr 2009 23:14:53 -0700 From: Ondrej Certik <ondrej(a)certik.cz> Subject: Re: [Numpy-discussion] DVCS at PyCon To: Discussion of Numerical Python <numpy-discussion(a)scipy.org> Message-ID: <85b5c3130904092314o7512cdbfjd2f43b98e65f48cf(a)mail.gmail.com> Content-Type: text/plain; charset=UTF-8 On Thu, Apr 9, 2009 at 10:45 PM, David Cournapeau <david(a)ar.media.kyoto-u.ac.jp> wrote: > Ondrej Certik wrote: >> >> It is maybe easier to learn how to work with different clones, but >> once you start working with lots of patches and you need to reclone >> all the time, then it's the wrong approach to work, as it takes lots >> of time to copy the whole repository on the disk. > > Yes, *I* know how to use git, and I agree with you, I vastly prefer git > branch handling to bzr branch handling. *I* find working with GUI for > VCS a real PITA. But I am not the only numpy developer, that's why the > feedback from people like Josef with a totally different workflow than > me is valuable - much more than people like us who are unix geeks :) Yes, definitely. Ondrej ------------------------------ Message: 2 Date: Fri, 10 Apr 2009 06:36:08 +0000 (UTC) From: Dan Lenski <dlenski(a)gmail.com> Subject: Re: [Numpy-discussion] using reducing functions without eliminating dimensions? To: numpy-discussion(a)scipy.org Message-ID: <grmpcn$2j4$1(a)ger.gmane.org> Content-Type: text/plain; charset=UTF-8 On Thu, 09 Apr 2009 01:31:33 -0500, Robert Kern wrote: > On Thu, Apr 9, 2009 at 01:29, Anne Archibald <peridot.faceted(a)gmail.com> > wrote: > >> What's wrong with np.amin(a,axis=-1)[...,np.newaxis]? > > It's cumbersome, particularly when you have axis=arbitrary_axis. Quite right. It would nice to be able to say: np.amin(a, axiskeep=-1) or a.min(axiskeep=3) ... or something along those lines. Dan ------------------------------ Message: 3 Date: Thu, 9 Apr 2009 23:42:03 -0700 From: Ian Mallett <geometrian(a)gmail.com> Subject: Re: [Numpy-discussion] Another Array To: Discussion of Numerical Python <numpy-discussion(a)scipy.org> Message-ID: <a62fab400904092342y5509976bk927277d72730a095(a)mail.gmail.com> Content-Type: text/plain; charset="iso-8859-1" It gives a perfect parabolic shape that looks very nice, but somewhat unrealistic. I'd like to scale the unit vectors by a random length (which can just be a uniform distribution). I tried scaling the unit vector n*n*3 array by a random n*n array, but that didn't work, obviously. Help?

1 0

numpy import on x86_64 arch
by Vincent Thierion 10 Apr '09

10 Apr '09

Hello, I import numpy module (numpy-1.0.4) on a x86_64 machine (on which I don't have any root privileges) after having install it thanks to python "setup.py install --prefix=../numpy". In this manner, I obtain a 64 bits compatible numpy library. (the "numpy" folder used for install is created just before install process) I had the following error : * import numpy File "../numpy/lib64/python2.3/site-packages/numpy/__init__.py", line 43, in ? File "../numpy/lib64/python2.3/site-packages/numpy/linalg/__init__.py", line 4, in ? File "../numpy/lib64/python2.3/site-packages/numpy/linalg/linalg.py", line 25, in ? ImportError: liblapack.so.3: cannot open shared object file: No such file or directory *I looked for liblapack.so.3 and find it in /usr/*lib*/liblapack.so.3. It seems it doesn't work because numpy 64 bits version seems to need /usr/* lib64*/liblapack.so.3. That I tested in other machine and it works when numpy can find /usr/*lib64*/liblapack.so.3. So is there a solution to use numpy 64 bits version with a /usr/*lib*/liblapack.so.3 or is it absolutly necessary to install lapack 64 bits version ? Thank you Vincent

2 1

frombuffer alignment for ctypes.Structure array
by Roland Schulz 10 Apr '09

10 Apr '09

Hi, I would like to create a ndarray view on an array of ctypes.Structure. (reason: see below). The problem is that the structure size is not falling on an alignment boundary and thus the size is bigger than its parts. A sample showing the problem: import numpy as N from ctypes import * class C(Structure): _fields_=[("a",c_int),("b",c_short)] c=(C*2)() _ctypes_to_numpy = {c_short : N.int16,c_int : N.int32} ty = [(x,_ctypes_to_numpy[y]) for x,y in C._fields_] x=N.frombuffer(c,ty) The is no align or aligned option to frombuffer. What is the best way to tell numpy to align the data as the C-struct/ctypes.Stucture array is? reason: I have an array which get allocated in the external C library and thus I get a pointer on structs from that C function. The only way I know how to create a ndarray view for pointers on data, is to first convert the ctypes pointer to a ctypes array. I convert the POINTER(t_X) to an ctypes array of the structre using the (t_X*N).from_address(addressof(X.contents)) trick, where t_X is the data type of the struct and X the pointer to the array. Thanks Roland -- ORNL/UT Center for Molecular Biophysics cmb.ornl.gov 865-241-1537, ORNL PO BOX 2008 MS6309

2 2

Another Array
by Ian Mallett 10 Apr '09

10 Apr '09

Hello, With the help of this list, over the past two days, I have implemented a GPU particle system (picture here: http://img7.imageshack.us/img7/7589/image1olu.png). Every particle is updated entirely on the GPU, with texture data (arrays) being updated iteratively between two framebuffer objects. Here, 65,536 particles are drawn at ~50 fps. Currently, jitter is introduced into the system by means of a "random texture", which should store random unit 3D vectors encoded into the RGB channels (vectors point in all directions, positive and negative, so the vecotr is actually 0.5 long and added to 0.5). The random texture is constructed using Python's random number generation, which is noticeably slower. I also didn't find an efficient way of creating "random" vectors. All this contributes to the pattern artifact at the bottom of the screenshot. I'd like to use NumPy. The array should be constructed as n*n*3, with every vec3 being a normalized random vector with each component in the range [-1.0,1.0]. I could use numpy.random.random(), but this would not give the appropriate vectors. Every 3D vector should have the same probability of being chosen. I don't know how to do that in NumPy--that's where I need help. Ian

3 9

using reducing functions without eliminating dimensions?
by Dan Lenski 10 Apr '09

10 Apr '09

Hi all, I often want to use some kind of dimension-reducing function (like min(), max(), sum(), mean()) on an array without actually removing the last dimension, so that I can then do operations broadcasting the reduced array back to the size of the full array. Full example: >> table.shape (47, 1814) >> table.min(axis=1).shape (47,) >> table - table.min(axis=1) ValueError: shape mismatch: objects cannot be broadcast to a single shape >> table - table.min(axis=1)[:, newaxis] I have to resort to ugly code with lots of stuff like "... axis=1)[:, newaxis]". Is there any way to get the reducing functions to leave a size-1 dummy dimension in place, to make this easier? Thanks! Dan

4 4

Numpy 1.3 on AMD64 import error
by Bruce Southey 09 Apr '09

09 Apr '09

Hi, I can not import numpy 1.3 on AMD64 on Windows Vista 64 business but it does import with the numpy-1.3.0b1. I get the first error below either via the command line or using IDLE. The second error occurs in IDLE if I try a second time to import numpy but probably not important. Bruce >>> import numpy Warning (from warnings module): File "C:\Python26\lib\site-packages\numpy\core\__init__.py", line 5 import multiarray Warning: Windows 64 bits support is experimental, and only available for testing. You are advised not to use it for production. CRASHES ARE TO BE EXPECTED - PLEASE REPORT THEM TO NUMPY DEVELOPERS Traceback (most recent call last): File "<pyshell#1>", line 1, in <module> import numpy File "C:\Python26\lib\site-packages\numpy\__init__.py", line 130, in <module> import add_newdocs File "C:\Python26\lib\site-packages\numpy\add_newdocs.py", line 9, in <module> from lib import add_newdoc File "C:\Python26\lib\site-packages\numpy\lib\__init__.py", line 13, in <module> from polynomial import * File "C:\Python26\lib\site-packages\numpy\lib\polynomial.py", line 18, in <module> from numpy.linalg import eigvals, lstsq File "C:\Python26\lib\site-packages\numpy\linalg\__init__.py", line 47, in <module> from linalg import * File "C:\Python26\lib\site-packages\numpy\linalg\linalg.py", line 22, in <module> from numpy.linalg import lapack_lite ImportError: DLL load failed: The specified module could not be found. >>> import numpy Traceback (most recent call last): File "<pyshell#2>", line 1, in <module> import numpy File "C:\Python26\lib\site-packages\numpy\__init__.py", line 130, in <module> import add_newdocs File "C:\Python26\lib\site-packages\numpy\add_newdocs.py", line 9, in <module> from lib import add_newdoc File "C:\Python26\lib\site-packages\numpy\lib\__init__.py", line 13, in <module> from polynomial import * File "C:\Python26\lib\site-packages\numpy\lib\polynomial.py", line 11, in <module> import numpy.core.numeric as NX AttributeError: 'module' object has no attribute 'core'

1 0

shift for optimal superimposition of two 3D matrices according to correlation computed using FFT
by Klaus Kopec 09 Apr '09

09 Apr '09

Hello everybody, I have two 3D density maps (meaning volumetric data, each one essentially a IxJxK matrix containing real values that sum to one) and want to find translation of between the two that maximises correlation. This can be done by computing the correlation between the two (correlation theorem -> I use FFT for this) and then shifting one of the two maps by the vector that corresponds to the index of the maximum in the 3D correlation matrix. My problem with this method is, however, that the shift wraps around the edges of the matrix. e.g: for a 30x30x30 matrix that needs to be slightly shifted to fit the seconds 3D matrix, sometimes shift vectors like (0, 29, 0) appear, that actually mean a shift by (0, -1, 0). This means that the size of the matrix needs to be subtracted from some components of the shift vector in special cases. So far, I have been unable to determine the cutoff value, outside of which I need to subtract the matrix size from a component of the shift vector, as this cutoff seems to vary between different data sets. Can anybody help me out with this? Best regards, Klaus

2 3

Specially Constructed Arrays
by Ian Mallett 09 Apr '09

09 Apr '09

Hello, I want to make an array of size sqrt(n) by sqrt(n) by 3, filled with special values. The values range from 0.0 to 3.0, starting with 0.0 at one corner and ending at 3.0 in the opposite, increasing going row by row. The value is to be encoded in each color. Because this is somewhat abstract, here's a small example (n=25), generated using the attached code (it also multiplies the number by 255 to obtain a RGB color and not messy floats) to show the concept. The real version should be done by NumPy. This is where I need help; I have no idea how to even approach the problem. [[ 0, 0, 0],[ 32, 0, 0],[ 64, 0, 0],[ 96, 0, 0],[128, 0, 0], [159, 0, 0],[191, 0, 0],[223, 0, 0],[255, 0, 0],[255, 32, 0], [255, 64, 0],[255, 96, 0],[255,128, 0],[255,159, 0],[255,191, 0], [255,223, 0],[255,255, 0],[255,255, 32],[255,255, 64],[255,255, 96], [255,255,128],[255,255,159],[255,255,191],[255,255,223],[255,255,255]] Arrays like this need to be generated quite quickly, so the per-pixel method I presented will not work. How should I do it with NumPy? Thanks, Ian

2 1