Mailman 3 February 2009 - NumPy-Discussion

Warnings in current trunk
by Charles R Harris Feb. 18, 2009

Feb. 18, 2009

I see a lot of warnings like: In file included from numpy/core/src/multiarraymodule.c:87: numpy/core/src/umath_funcs_c99.inc.src:199:1: warning: "isnan" redefined In file included from /usr/include/python2.5/pyport.h:204, from /usr/include/python2.5/Python.h:57, from numpy/core/src/multiarraymodule.c:18: These are probably from r6363. Also In file included from numpy/core/src/scalartypes.inc.src:10, from numpy/core/src/arrayobject.c:537, from numpy/core/src/multiarraymodule.c:102: numpy/core/src/numpyos.c: In function 'NumPyOS_ascii_strtod': numpy/core/src/numpyos.c:431: warning: assignment discards qualifiers from pointer target type numpy/core/src/numpyos.c:480: warning: assignment discards qualifiers from pointer target type Chuck

2 8

views and object lifetime
by Neal Becker Feb. 18, 2009

Feb. 18, 2009

How is it ensured, at the C api level, that when I have an array A, and a view of it B, that the data is not destroyed until both A and B are?

5 6

Re: [Numpy-discussion] Compute multiple outer products, without a loop?
by Ken Basye Feb. 18, 2009

Feb. 18, 2009

Thanks Chuck; that's perfect. Ken > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 17 Feb 2009 11:04:56 -0700 > From: Charles R Harris <charlesr.harris(a)gmail.com> > Subject: Re: [Numpy-discussion] Compute multiple outer products > without a loop? > To: Discussion of Numerical Python <numpy-discussion(a)scipy.org> > Message-ID: > <e06186140902171004g850c9aax348a3e2053bb69fa(a)mail.gmail.com> > Content-Type: text/plain; charset="iso-8859-1" > > On Tue, Feb 17, 2009 at 8:30 AM, Ken Basye <kbasye1(a)jhu.edu> wrote: > > >> Hi, >> My current code looks like this: >> >> (k,d) = m.shape >> sq = np.zeros((k, d, d), dtype=float) >> for i in xrange(k): >> sq[i] = np.outer(m[i], m[i]) >> >> >> That is, m is treated as a sequence of k vectors of length d; the k dXd >> outer products are found and stored in sq. >> >> > > Try A[:,:,newaxis]*B[:,newaxis,:] . Example > > In [6]: A = array([[1,2],[3,4]]) > > In [7]: B = array([[1,1],[1,1]]) > > In [8]: A[:,:,newaxis]*B[:,newaxis,:] > Out[8]: > array([[[1, 1], > [2, 2]], > > [[3, 3], > [4, 4]]]) > > In [9]: B[:,:,newaxis]*A[:,newaxis,:] > Out[9]: > array([[[1, 2], > [1, 2]], > > [[3, 4], > [3, 4]]]) > > You can use this sort of trick along with a sum to multiply stacks of > matrices by stacks of vectors or matrices. > > Chuck > >

1 0

fancy view question
by Vincent Schut Feb. 18, 2009

Feb. 18, 2009

Hi list, would it be possible to create a view on an array, such that this view is twice as large (in some dimensions) and in fact does a nearest neighbour 'zoom' on the original array? E.g. using some fancy slicing/striding tricks? an example: a = [[1, 2], [3, 4]] then I'd like a view on a such that this view is: [[1, 1, 2, 2], [1, 1, 2, 2], [3, 3, 4, 4], [3, 3, 4, 4]] I know several ways to create this as a copy, but as memory does play a role in this case (yes, I know it's cheap :-))... If not, fancy ways to create this as a copy are apreciated. I currently either create an empty array and then loop-fill it (e.g. this would be 2x2 loops), of if I' lazy I use ndimage.zoom, which actually is a bit overkill. Regards, Vincent.

6 19

Tensordot and memory consumption
by Nicolas Rougier Feb. 18, 2009

Feb. 18, 2009

Hello, I'm using tensordot in some computation and while I've been amazed by the speed, I'm now trying to reduce memory consumption in some very particular cases: Let S be a 2 dims array of size (s1,s2) Let D be a 2 dims array of size (d1,d2) Let W be a 4 dims array of size (d1,d2,s1,s2) Currently, I'm computing D as tensordot(W,S,2) and it works really fine and fast. However, in some cases, W is based solely on a single 2 dims array K and each of the W[i,j] is a slice of K. I would like to know if there is a way to build W such that memory footprint is reduced ? Or maybe there is other ways to perform the same computation ? For example, I know that when S and D are of same shape, I can use 2d convolution, but I would need the general case. Since I imagine my explanations are not sot clear, I join a small example. Nicolas --- import numpy n = 3 # Source S = numpy.random.random((n,n)) # Destination (at this stage, only D shape matters) D = numpy.zeros((2*n,2*n)) # Kernel K = numpy.zeros((S.shape[0]*n+1,S.shape[1]*n+1)) K[S.shape[0],S.shape[1]] = 1 # Kernel decomposition for computing tensordot W = numpy.zeros(D.shape + S.shape) for i in range(W.shape[0]): for j in range(W.shape[1]): x = int(i/float(W.shape[0])*float(S.shape[0])) y = int(j/float(W.shape[1])*float(S.shape[1])) W[i,j] = K[n-x:n-x+S.shape[0],n-y:n-y+S.shape[1]] D = numpy.tensordot(W,S,2) print S print print D

1 0

lack of memory?
by Xiaoyu Chu Feb. 18, 2009

Feb. 18, 2009

Hello Everyone, I am trying to compute the eigenvalues of a matrix of size 200*200. What happened is that whenever I run the program, python dies, without showing an error message. The program works fine when the matrix size is small, like in 10s. Does anyone know if it is because of the memory issue or something else? By the way, I am using the method linalg.eigvals(M) Many thanks. -- Best Regards, Xiaoyu

4 3

Re: [Numpy-discussion] Compute multiple outer products without a loop?
by Ken Basye Feb. 17, 2009

Feb. 17, 2009

Hi, My current code looks like this: (k,d) = m.shape sq = np.zeros((k, d, d), dtype=float) for i in xrange(k): sq[i] = np.outer(m[i], m[i]) That is, m is treated as a sequence of k vectors of length d; the k dXd outer products are found and stored in sq. Note that the expression np.outer(m, m) will always produce an array with shape (p, p) where p is the product over the shape of m. E.g. if m.shape == (10, 4) then outer(m, m).shape == (40, 40) whereas I want something with shape (10, 4, 4). I suppose all the values I want are indeed somewhere in this 2d array, since it contains every pairwise product, but I'm not sure I want to do that much extra computation just to avoid the loop, plus I'd still have to figure out how to get a view that covered only the results I care about. Thanks, Ken

2 1

efficient function calculation between two matrices
by John [H2O] Feb. 17, 2009

Feb. 17, 2009

hello, I am trying to calculate the results of a function between two matrices: >>> F.shape (170, 2) >>> T.shape (170, 481, 2) Where F contains lat/lon pairs and T contains 481 lat/lon pairs for 170 trajectories of length 481 I want a new array of shape (170,481) containing the results of my distance function: gcd(x1,y1,x2,y3): """ returns great circle distance""" ... return dist Is there a way to do this without the use of loops? I'm thinking something along the lines of: distances = [[gcd(f[0],f[1],t[0],t[1]]) for f in F] for t in T] But obviously this doesn't work... Thank you, john -- View this message in context: http://www.nabble.com/efficient-function-calculation-between-two-matrices-t… Sent from the Numpy-discussion mailing list archive at Nabble.com.

2 1

GMRES internal variables
by jordan＠math.ucsb.edu Feb. 17, 2009

Feb. 17, 2009

Hi all, I'm trying to run multiple instances of GMRES at the same time (one inside another actually, used inside of the preconditioner routine) but am running in to some problems. My guess is there is a single set of internal variables associated with GMRES and when I initiate a new GMRES inside a currently running GMRES, the old GMRES data gets the boot. Is this correct, and if so any suggestions on how to fix this? Cheers, Jordan

3 2

Compute multiple outer products without a loop?
by Ken Basye Feb. 16, 2009

Feb. 16, 2009

Hi List, I need to compute multiple outer products from 2-d data in the following way: Given a and b with shape, e.g, (10, 4), compute the 10 outer products of shape (4, 4) and get them into an array of shape (10, 4, 4). Currently I do this with a loop, but I'd really like some way to do it without looping. I read that outer(a, b) is just syntactic sugar for a.ravel()[:, newaxis] * b.ravel()[newaxis,:] but unfortunately this didn't give me any bright ideas :->. Thanks, Ken

2 1