Re: [Numpy-discussion] expensive tensordot

Dag Sverre Seljebotn student.matnat.uio.no> writes: > > Paul Northug wrote: > > I have a computation bounded by one step and I have always wondered > > how to make it fast enough to be useable. I suspect that I have to use > > an approximation, but I was hoping

[Numpy-discussion] expensive tensordot

I have a computation bounded by one step and I have always wondered how to make it fast enough to be useable. I suspect that I have to use an approximation, but I was hoping someone would spot a major inefficiency in my implementation. The calculation is a kind of outer product of two sets of time

Re: [Numpy-discussion] reformulating a series of correlatio ns as one fft, ifft pair

Stéfan van der Walt sun.ac.za> writes: > > I haven't checked your code in detail, but I'll mention two common > problems with this approach in case it fits: > > 1) When convolving a KxL with an MxN array, they both need to be zero > padded to (K+M-1)x(L+N-1). > 2) One of the signals needs to be

Re: [Numpy-discussion] reformulating a series of correlations as one fft, ifft pair

gmail.com> writes: > > On Mon, Apr 19, 2010 at 2:36 AM, Paul Northug gmail.com> wrote: > > > > # > > import numpy as np > > from scipy.signal import fftn, ifftn, correlate > > > > M, N, P, T = 2, 3, 4, 5 > > > > phi = np.random.randn

[Numpy-discussion] reformulating a series of correlations as one fft, ifft pair

I am having trouble reformulating a series of correlations as a single fft, ifft pair. I have a set of kernels phi : (M = channel, N = kernel, T = time) correlated with signal a : (N, P+T-1) yielding x : (M, T). The correlation, for now, is only in the last dimension, with the two other dimension

Re: [Numpy-discussion] Fortran ordering

Stéfan van der Walt sun.ac.za> writes: > > On 16 April 2010 21:35, Paul Northug gmail.com> wrote: > > how is it stored in memory, as [1, 2, 3, 4] or [1, 3, 2, 4]? > > The latter: > > In [22]: np.fromstring(str(x.data)) > Out[22]: array([ 1., 3., 2., 4.]

[Numpy-discussion] Fortran ordering

I'd like to use numpy fortran ordering in order to use some external libraries more easily, but it looks like I don't understand how it works and it is causing transposes that are confusing me. When I define an array as: a = np.array([[1.,2.],[3.,4.]], order='F', type=np.float32) how is it store

[Numpy-discussion] memoization with ndarray arguments

I would like to 'memoize' the objective, derivative and hessian functions, each taking a 1d double ndarray argument X, that are passed as arguments to scipy.optimize.fmin_ncg. Each of these 3 functions has calculations in common that are expensive to compute and are a function of X. It seems fmin_