On 10-Jul-09, at 1:25 AM, Chris Colbert wrote:
actually what would be better is if i can pass two 1d arrays X and Y
both size Nx1
and get back a 2d array of size NxM where the [n,:] row is the linear
interpolation of X[n] to Y[n]
This could be more efficient, but here's a solution using
On 10-Jul-09, at 1:26 PM, David Goldsmith wrote:
grid = np.array([np.linspace(x[i],y[i],nrows) for i in
range(len(x))]).T
Indeed, linspace will work, but careful with Python loops though,
it'll be 2x to 6x slower (based on my empirical fiddling) than the
solution involving mgrid.
In
On 9-Jul-09, at 1:12 AM, Mag Gam wrote:
Here is what I have, which does it 1x1:
z={} #dictionary
r=csv.reader(file)
for i,row in enumerate(r):
p=/MIT/+row[1]
if p not in z:
z[p]=0:
else:
z[p]+=1
arr[p]['chem'][z[p]]=tuple(row) #this loads the array 1 x 1
I would like to
On 8-Jul-09, at 6:16 PM, Pauli Virtanen wrote:
Just to tickle some interest, a pathological case before
optimization:
In [1]: import numpy as np
In [2]: x = np.zeros((8, 256))
In [3]: %timeit x.sum(axis=0)
10 loops, best of 3: 850 ms per loop
After optimization:
In
That is some weird mojo; I'm guessing that numpy simply doesn't
recognize the 'LA' type (luminance+alpha) as it's pretty uncommon. In
the meantime you probably want to convert to RGBA, since that file
giving you problems is grayscale+alpha channel, and a conversion to
RGB might lose the
On 21-Jun-09, at 11:59 AM, David Cournapeau wrote:
Can't really say at this point, but it is the suggested path to
python-3.
OTOH, I don't find the python 3 official transition story very
convincing. I have tried to gather all the information I could find,
both on the python wiki and from
On 17-Jun-09, at 2:18 PM, Nils Wagner wrote:
Is there a port of numpy/scipy to Jython ?
Any pointer would be appreciated.
Folks have successfully gotten it working from IronPython (the .NET
CLR) via Ironclad ( http://code.google.com/p/ironclad/ )... not Jython
though.
David
On 10-Jun-09, at 3:23 PM, Gökhan SEVER wrote:
I am very off-the-topic, sorry about that first, but I know most of
the
people in this list are students / scientists. Just want to know a few
opinions upon how you manage references (following the list of
references in
the end of articles,
On 9-Jun-09, at 3:54 AM, David Cournapeau wrote:
For example, what ML people call PCA is called Karhunen Loéve in
signal
processing, and the concepts are quite similar.
Yup. This seems to be a nice set of review notes:
http://www.ece.rutgers.edu/~orfanidi/ece525/svd.pdf
And going
On 9-Jun-09, at 2:56 PM, bruno Piguet wrote:
Phi is now of size(n) and V (n, 3).
(I really whish to have this shape, for direct correspondance to
file).
The corresponding function looks like :
def rotat_vect(phi, V):
s = np.sin(phi)
c = np.cos(phi)
M = np.zeros((len(phi), 3,
On 7-Jun-09, at 4:56 AM, giorgio.luci...@inwind.it wrote:
Sorry for cross posting
Hello to all,
I've done a script for importing all spectra files in a directory
and merge all them in one matrix. The file imported are dx files.
the bad part is that the file is in matlab and it requite a
On 8-Jun-09, at 1:17 AM, David Cournapeau wrote:
I would not be surprised if David had this paper in mind :)
http://www.cs.toronto.edu/~roweis/papers/empca.pdf
Right you are :)
There is a slight trick to it, though, in that it won't produce an
orthogonal basis on its own, just something
On 8-Jun-09, at 8:33 AM, Jason Rennie wrote:Note that EM can be very slow to converge:That's absolutely true, but EM for PCA can be a life saver in cases where diagonalizing (or even computing) the full covariance matrix is not a realistic option. Diagonalization can be a lot of wasted effort if
On 8-Jun-09, at 12:58 PM, Jonno wrote:
Thanks Josef,
I shouldn't have included Matplotlib since Pydee does work well with
its plots. I had forgotten that. It really is just the Mayavi plots
(or scenes I guess) that don't play well.
I don't know how exactly matplotlib integration issues are
A question was raised on the #scipy IRC earlier today, about the
behaviour of array() with structured dtypes. After some educated
guessing I figured out that for record arrays, tuples (rather than
lists) must be used to indicate atomic elements. What I wondered is
whether this behaviour is
On 7-Jun-09, at 6:12 AM, Gael Varoquaux wrote:
Well, I do bootstrapping of PCAs, that is SVDs. I can tell you, it
makes
a big difference, especially since I have 8 cores.
Just curious Gael: how many PC's are you retaining? Have you tried
iterative methods (i.e. the EM algorithm for PCA)?
On 4-Jun-09, at 4:38 PM, Anne Archibald wrote:
It seems to me that this is the basic source of the problem. Perhaps
this can be addressed? I realize maintaining compatibility with the
current behaviour is necessary, so how about a multistage deprecation:
1. add a keyword argument to
On 3-Jun-09, at 5:01 PM, Pauli Virtanen wrote:
Btw, are you able to change the status of the ticket to
needs_review?
I think this should be possible for everyone, and not restricted to
admins, but I'm not 100% sure...
Sorry Pauli, seems I _don't_ have permission on the numpy trac to
On 4-Jun-09, at 9:28 AM, Stéfan van der Walt wrote:
2009/6/4 David Warde-Farley d...@cs.toronto.edu:
Sorry Pauli, seems I _don't_ have permission on the numpy trac to
change ticket status. The radio button shows up but then it gives
me a
Warning: No permission to change ticket fields
On 4-Jun-09, at 5:03 PM, Anne Archibald wrote:
Apart from the implementation issues people have chimed in about
already, it's worth noting that the speed of matrix multiplication
depends on the memory layout of the matrices. So generating B instead
directly as a 100 by 500 matrix might affect
On 2-Jun-09, at 3:06 PM, Pauli Virtanen wrote:
+0
I don't see any drawbacks, and the implementation looks good.
Thanks Pauli. I realized I was missing values() and itervalues()
(though I can't conceive of a scenario where I'd use them myself, I
guess some code might expect them). Also I
On 3-Jun-09, at 5:01 PM, Pauli Virtanen wrote:
Btw, are you able to change the status of the ticket to
needs_review?
I think this should be possible for everyone, and not restricted to
admins, but I'm not 100% sure...
Sorry, yes I am. I had just forgotten.
David
Hi,
It's occasionally annoyed me that NpzFiles can't be swapped in
transparently for an in-memory dictionary since getting at the keys
requires an attribute access. Below is a patch that implements some
more of the dictionary interface for the NpzFile class. Any comments
as to whether
On 26-May-09, at 1:15 AM, Robert Kern wrote:
I *did* do some due diligence before I designed a new binary format.
Uh oh, I feel this might've taken a sharp turn towards another of
course Robert is right, Robert is always right threads. :)
David
On 24-May-09, at 5:22 PM, Robert Kern wrote:
While I haven't tried Andrew Collette's h5py
(http://code.google.com/p/h5py), it looks like a very 'thin' wrapper
around the HDF5 C libraries. Maybe numpy's save(), savez(), load(),
memmap() could be enhanced so that saving/loading files with
On 24-May-09, at 8:32 AM, Tom K. wrote:
Maybe my reluctance to work with matrices stems from this kind of
inconsistency. It seems like your code has to be all matrix, or all
array -
and if you mix them, you need to be very careful about which is which.
Also, functions called on things of
On 23-May-09, at 5:36 AM, Albert Thuswaldner wrote:
So i guess in the long term i have to also add pickling support. In
the short term i will add warnings for the data types that are not
supported.
In order to ensure optimal division of labour, I'd suggest simply
basing your pickling
Can someone with the requisite permissions change the title of ticket
#1113 to reflect the fact that it affects both ppc and ppc64?
Alternately, if you know why the bug is happening, you could file a
patch ;)
David
___
Numpy-discussion mailing list
On 23-May-09, at 8:54 AM, David Cournapeau wrote:
I have not looked at the code, but if the precision is indeed single
precision, a tolerance of 1e-15 may not make much sense (single
precision has 7 significant digits in normal representation)
Yes, I was wondering about that too, though
On 23-May-09, at 2:33 PM, Pauli Virtanen wrote:
Yes, I was wondering about that too, though notably the tests pass on
x86, and in fact the result on ppc was nowhere near 0 when I
checked it.
What do you mean by nowhere near? What does the following output for
you:
On 23-May-09, at 4:25 PM, Albert Thuswaldner wrote:
Actually my vision with pyhdf5io is to have hdf5 to replace numpy's
own binary file format (.npy, npz). Pyhdf5io (or an incarnation of it)
should be the standard (binary) way to store data in scipy/numpy. A
bold statement, I know, but I
On 23-May-09, at 4:59 PM, Robert Kern wrote:
Otherwise it will just work the way it does now.
That would cause difficulties. Now the format of your data depends on
whether or not you have a package installed. That's not a very good
level of control.
Sorry, I wasn't clear. What I meant
On 22-May-09, at 1:03 PM, Christopher Barker wrote:
In [104]: zip(indices[np.r_[True, breaks[:-1]]], indices[breaks])
I don't think this is very general:
In [53]: indices
Out[53]:
array([ -3, 1, 2, 3, 4, 5, 6, 7, 8,
9, 255, 256, 257, 258,
Hi Albert,
So this is a wrapper on top of PyTables to implement load() and
save()? Neat.
Obviously if you're installing PyTables, you can do a lot better and
organize your data hierarchically without the messiness of Matlab
structures, walk the node tree, all kinds of fun stuff, but if
I've just tried a fat 64 build (with a Python 2.6.2 that had been
built similarly), and I'm getting this weird behaviour. The command I
used was:
CFLAGS=-O3 -Wall -DNDEBUG -g -fwrapv -Wstrict-prototypes -arch x86_64
-arch ppc64 python setup.py build
It looks as though for some reason,
On 11-May-09, at 10:55 AM, Pauli Virtanen wrote:
Wonder why buildbot's 64-bit SPARC boxes don't see this if it's
something
connected to 64-bitness...
Different endianness, maybe? That seems even weirder, honestly.
David
___
Numpy-discussion
On 12-May-09, at 3:55 PM, Ryan May wrote:
It's going to be faster to do it without the transpose. Besides,
for numpy,
that imshow becomes:
imshow(b[0])
Which, IMHO, looks better than Matlab.
You're right, that is better, odd how I never thought of doing it like
that. I've been
On 6-May-09, at 2:03 AM, Christopher Barker wrote:
maybe:
numpy-1.3.0-py2.5-macosx-python.org.dmg
+1 on having python.org in the name. It clarifies and reinforces the
case that this isn't for the Apple-shipped Python (which I heard
comes with NumPy now?).
David
Hi,
Is there a simple way to compare each element of an object array to a
single object? objarray == None, for example, gives me a single
False. I couldn't find any reference to it in the documentation, but
I'll admit, I wasn't quite sure where to look.
David
On 28-Apr-09, at 10:56 AM, Dan Goodman wrote:
Can anyone explain the results below? It seems that for small matrices
dot(x,y) is outperforming dgemm(1,x,y,0,y,overwrite_c=1), but for
larger
matrices the latter is winning. In principle it seems like I ought
to be
able to always do better
On 29-Apr-09, at 5:06 PM, Dan Goodman wrote:
Here's the problem I want to write vectorised code for. I start with
an
array of indices, say I=array([0,1,0,2,0,1,4]), and I want to come up
with an array C that counts how many times each index has been seen so
far if you were counting
On 29-Apr-09, at 5:49 PM, Dan Goodman wrote:
Thanks David, that's nice but unfortunately that Python loop will kill
me. I'm thinking about some simulation code I'm writing where this
operation will be carried out many, many times, with large arrays I. I
figure I need to keep the Python
On 9-Jan-09, at 4:31 PM, Robert Kern wrote:
You can't in numpy. With scipy.linalg.fblas.dgemm() and the right
arguments, you can.
Make sure your output array is Fortran-ordered, however, otherwise
copies will be made.
David
___
Numpy-discussion
On 24-Apr-09, at 10:11 PM, Chris Colbert wrote:
Like the subject says, is there a way to register numpy with
synaptic after building numpy from source?
I would like to snag matplotlib from the ubuntu repos, but it won't
let me without also getting numpy and its dependency, which would
Alan G Isaac wrote:
On 3/27/2009 6:48 AM David Cournapeau apparently wrote:
To build the numpy .dmg mac os x installer, I use a script from the
adium project, which uses applescript and some mac os x black magic. The
script seems to be GPL, as adium itself:
It might be worth a query
Hi all,
I built ATLAS, Python 2.5 and NumPy on the local disk of a cluster
node, so that disk access would be faster than over NFS, and then
moved it back. I made sure to modify all the relevant paths in
__config__.py but when importing I receive this error, which I can't
make heads or
On 26-Mar-09, at 3:32 PM, Ben Park wrote:
BTW, this timing on a core 2 Duo 2.0GH laptop ,with the Enthought
Python
Distribution, is around 0.2 second.
You're going to have to build NumPy yourself to link it against the
MKL, I believe. EPD's is probably using something fairly basic.
You
On 20-Feb-09, at 6:41 AM, Olivier Grisel wrote:
Alright, thanks for the reply.
Is there a canonical way /sample code to gain low level access to
blas / lapack
atlas routines using ctypes from numpy / scipy code?
I don't mind fixing the dimensions and the ndtype of my array if it
can
On 3-Mar-09, at 11:41 PM, Jonathan Taylor wrote:
def rotation(theta, R = np.zeros((3,3))):
Hey Jon,
Just a note, in case you haven't heard this schpiel before: be careful
when you use mutables as default arguments. It can lead to unexpected
behaviour down the line.
The reason is that the
On 4-Mar-09, at 1:58 AM, Robert Kern wrote:
I'm pretty sure that's exactly why he did it, and that's what he's
calling evil.
As ever, such nuance is lost on me. I didn't bother to check whether
or not it was in the original function. Robert to the rescue. :)
It's a neat trick, actually,
On 4-Mar-09, at 1:50 AM, Hoyt Koepke wrote:
I would definitely encourage you to check out cython. I have to write
lots of numerically intensive stuff in my python code, and I tend to
cythonize it a lot.
Seconded. I recently took some distance computation code and
Cythonized it, I got an
On 28-Feb-09, at 12:27 PM, Jonathan Taylor wrote:
This does seem like the only way to write this nicely. Unfortunately,
I think this may be wasteful memory wise (in contrast to what the
obvious matlab code would do) as it constructs an array with the whole
first index intact at first.
True
On 2-Mar-09, at 12:25 PM, Robert Kern wrote:
a[[2,3,6], ...][..., [3,2]]
You're doing fancy indexing, so there are copies both times.
D'oh!
So I guess the only way to avoid the second copy is to do what Jon
initially suggested, i.e. a[ix_([2,3,6],range(a.shape[1]),[3,2])] ?
I suppose
Hey Jon,
On 26-Feb-09, at 10:00 PM, Jonathan Taylor wrote:
Am I right to assume that there is no way elegant way to interact with
slices. i.e. Is there anyway to get
a[ix_([2,3,6],:,[3,2])]
to work? So that the dimension is completely specified? Or perhaps
the only way to do this is
On 27-Feb-09, at 3:35 PM, David Warde-Farley wrote:
a[[2,3,6],:,:][:,:,[3,2]] should do what you want.
Slightly more elegantly (I always forget about this syntax):
a[[2,3,6], ...][..., [3,2]]
David
___
Numpy-discussion mailing list
Numpy-discussion
Hi Olivier,
There was this idea posted on the Scipy-user list a while back:
http://projects.scipy.org/pipermail/scipy-user/2008-August/017954.html
but it doesn't look like he got anywhere with it, or even got a
response.
I just tried it and I observe the same behaviour. A quick look
On 20-Feb-09, at 6:39 AM, David Cournapeau wrote:
You can just use ctypes to access ATLAS, as you would do for any
library. Or do you mean something else ?
Say, David... :)
Do you have any idea why the pyf wrapper for fblas3 completely ignores
the overwrite_c argument? Fiddling around I've
On 20-Feb-09, at 10:39 AM, Robert Kern wrote:
Fiddling around I've found other blas/lapack
functions where the same arg is offered but the choice actually does
something.
Examples?
scipy.lib.lapack.flapack.dpotri, for example. I'm not sure of the
proper usage, but when I pass it an
On 20-Feb-09, at 10:39 AM, Robert Kern wrote:
Fiddling around I've found other blas/lapack
functions where the same arg is offered but the choice actually does
something.
Examples?
An even better example is scipy.linalg.fblas.dgemv, the matrix-vector
equivalent of dgemm. overwrite_y
On 28-Nov-08, at 5:38 PM, Gideon Simpson wrote:
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?
IIRC David Cournapeau has had some success building numpy with MKL on
On 24-Nov-08, at 4:22 PM, Gabriel Gellner wrote:
asin(1j) raises an exception, arcsin doesn't. They are *different*
functions, hence the names.
Yet:
type(np.sin(1)) == type(math.sin(1))
False
In fact, this goes for every single function listed in the math
module's docs, except for the
On 18-Nov-08, at 5:29 AM, Nicolas ROUX wrote:
Hi,
Maybe this is not so clever, but I can't find it in the doc.
I need to get all indices/index of all occurrences of a value in a
numpy
array
As example:
a = numpy.array([1,2,3],[4,5,6],[7,8,9])
I need to get the indice/index of all
On 12-Nov-08, at 8:18 PM, David Cournapeau wrote:
On Wed, 2008-11-12 at 19:24 -0500, David Warde-Farley wrote:
Indeed, for the size of problem I *thought* I was running, 32 bit
would be sufficient. In fact I had my data transposed and so was
working with a much larger matrix which would
On 13-Nov-08, at 8:47 PM, David Cournapeau wrote:
On Fri, Nov 14, 2008 at 5:23 AM, frank wang [EMAIL PROTECTED] wrote:
Hi,
Can you provide a working example to build Numpy with MKL in window
and
linux?
The reason I am thinking to build the system is that I need to make
the
speed
Hello folks,
I'm doing some rather big matrix products on a G5, and ran into this.
Strangely on the same OS version on my Intel laptop, this isn't an
issue. Available memory isn't the problem either, I don't think, this
machine is pretty beefy.
I'm running the python.org 2.5.2 build of
On 12-Nov-08, at 6:05 PM, Michael Abshoff wrote:
I'm running the python.org 2.5.2 build of Python, and the latest SVN
build of numpy (though the same thing happened with 1.1.0).
IIRC that is a universal build for 32 bit PPC and Intel, so
depending on
the problem size 32 bits might be
On 6-Nov-08, at 11:15 PM, Angus McMorland wrote:
2008/11/6 Robert Kern [EMAIL PROTECTED]:
On Thu, Nov 6, 2008 at 21:54, Angus McMorland [EMAIL PROTECTED]
wrote:
Hi all,
I'm trying to import a 16-bit tiff image into a numpy array. I have
found, using google, suggestions to do the
On 29-Oct-08, at 3:43 PM, Robert Kern wrote:
Eh, that's not entirely true.
x = 1
x += 2
That's not in-place. They are called augmented assignments, not
in-place operations for this reason. The defining characteristic is
that x op= y should be equivalent to x = x op y except
possibly
On 28-Oct-08, at 5:57 PM, Fabrice Silva wrote:
Are there some parts of the code that may be used only once to
calculate
both the gradient and the second derivative (isn't it called the
hessian, at least in the N-d case) ?
Probably. I'd imagine depends on your differencing scheme; central
On 27-Oct-08, at 7:22 AM, James Philbin wrote:
One operator which could be used is '%'. We could keep the current
behaviour for ARRAY%SCALAR but have ARRAY%ARRAY as being matrix
multiplication. It has the same precedence as *,/.
The problem is that it would monkey with existing semantics for
On 26-Oct-08, at 9:43 AM, James Philbin wrote:
This hack for defining infix operators might be relevant:
http://code.activestate.com/recipes/384122/
I think someone mentioned this at the doc BOF, but it was raised that
this has problems with associativity, etc.
David
201 - 271 of 271 matches
Mail list logo
