Mailman 3 November 2007 - SciPy-User

Kernel density approximation
by Gael Varoquaux 05 Nov '07

05 Nov '07

Hi, What is the recommended way to compute a kernel density approximation of a probability distribution function given a sample data ? My problem is two dimensional, each data element is made of a pair of numbers. I currently use an histogram2d, but I would like to use a smooth kernel, such as a gaussian, and most important, I would like the algorithm to do the work for me as far as choosing the bandwidth :->. And I forgot to say that the function should be fast, as it is used in a Monte Carlo algorithm, and is called many times. Yes, I know, I want the Moon, on a stick, please :->. Thanks, Gaël

2 2

Re: [SciPy-user] Smoothing of a signal
by Gael Varoquaux 05 Nov '07

05 Nov '07

Hi, On Mon, Nov 05, 2007 at 07:20:35AM -0800, Marc Beck wrote: > I am a graduate assistant at the Space Science Center at Morehead State > University. I am currently writing a program to convert raw data from our > 21m antenna into a *.FITS image using numpy and scipy. I need to convolve > the pixels of my image. I have three arrays: one for the x- coordinate, > one for the y- coordinate and one for the brightness of the pixel. I need > to feed those three arrays into a function to convolve the pixels and I > need to get the same three arrays out with the new values, or the rest of > the program does not work. Someone from the python forum referred me to > some code you wrote at [1]http://www.scipy.org/Cookbook/SignalSmooth > but I don't understand which variable is what and how to integrate this > into my program. Please explain to me how I can make my program work. I > hope that I have described clearly enough what I am planning to do. Your are better off asking the mailing list for these kind of question. I am terribly busy currently (important deadlines) and almost deleted you message before realising it had been sent only to me. Is you data on a regular grid ? You currently have an (X, Y, V) data layout, and it would be much easier to have a 2D array as a layout for your data, where the row and column index would give the X and Y position (scaled so that they fit in consecutive integers). With some luck you can convert you V array to a properly-shaped 2D array simply using "reshape". This correpsonds to the Z array in the code on the page you reference. Hope this helps, Gaël

1 0

surface plotting
by Joe Harrington 04 Nov '07

04 Nov '07

I'm in search of either Python or an open-source, standalone image viewer that can visualize surface plots. A surface plot is a rendering of a 2D image as a 3D surface using the image values as the Z coordinate. The new wiremesh in matplotlib is way too slow to be of practical use. It takes over a minute to plot a 1024**2 image, and for my purposes I need to move the surface around and look at it from different angles, which requires many re-renderings. Previously, I had used IDL, which has two routines called surface and shade_surface. The first is a wiremesh, the second is the same but adds shading according to viewing angle. These routines ran at decent speeds in IDL even when they were introduced about 20 years ago. IRAF also has them (or at least the wiremesh), but its size and klunkiness make it a poor solution as well. I'm not sure what's up with matplotlib, but this is a very new feature and it just might not be doing its thing in the most efficient way. In the long run, perhaps the source code for the IRAF routine could be a guide for doing it efficiently. In the short run, any ideas? It would make a good cookbook recipe. --jh--

6 6

[ANN] Scope_Plot, another plot library for real time signals.
by Stef Mientki 04 Nov '07

04 Nov '07

hello, I justed finished, another plot library, called Scope_Plot, based on wxPython. Scope_Plot is special meant for displaying real time signals, and therefor has some new functionalities: - signal selection - each signal has it's own scale, - moving erase block - measurement cursor ans should be a lot faster than MatPlot, Plot and FloatCanvas, at least for real time signals (because it only draws the changes). An animated demo can be seen here (2 MB, 2:10): http://stef.mientki.googlepages.com/jalspy_scope.html A description of the library, as used in an application, can be found here: http://oase.uci.kun.nl/~mientki/data_www/pic/jalspy/jalspy_scope.html And finally the file (and a few necessary libs) can be found here: http://oase.uci.kun.nl/~mientki/download/Scope_Plot.zip The library has a main section which contains a simple demo, the animated demo and the application description shows a more complex signal organization. cheers, Stef Mientki

1 0

Re: [SciPy-user] Subject: Re: Scientific Python publications
by Nicolas Pettiaux 04 Nov '07

04 Nov '07

2007/10/28, Tony Mannucci <Tony.Mannucci(a)jpl.nasa.gov>: > >From: jaropis <jaropis(a)gazeta.pl> > >I use Python a lot - for example this paper: > > > >http://www.iop.org/EJ/abstract/0967-3334/28/3/005 > > > >has an animation done in SciPy+pylab, and all the calculations are in > >Python. I give Matlab code for the same calculations in an appendix, but > >they were originally done in SciPy and then translated to Matlab. Maybe Jarek could send us or post on his website a copy of a previous version of the paper. These are not covered usually by the (abusive) copyright of the Journal THanks, Nicolas -- Nicolas Pettiaux - email: nicolas.pettiaux(a)ael.be Utiliser des formats ouverts et des logiciels libres - http://www.passeralinux.org. Pour la bureautique, les seuls formats ISO sont ceux de http://fr.openoffice.org

9 17

Re: [SciPy-user] Scipy + Vision = LabView
by Jose Unpingco 04 Nov '07

04 Nov '07

All: I put the tutorial together, and I plan to do more along these lines. My interest is in using VISION as a framework for a detailed communications simulation, although up to this point, VISION has been primarily used as a tool in chemistry and biology. I plan to write certain communications and signal processing components based upon scipy and integrate them into VISION as a distinct block set (i.e. component library). The developers of VISION are situated nearby and I am looking forward to that collaboration. Progress will be announced on the VISION mailing list (http://mgldev.scripps.edu/mailman/listinfo/vision) and on my blog (http://www.osc.edu/blogs/index.php/sip), but not on this list to avoid polluting it. As always, I appreciate your criticism. --José

1 0

resizing a dok_matrix
by Evan Lezar 04 Nov '07

04 Nov '07

Hi I have been using a dok_matrix from scipy 0.5.3 to implement re-sizable sparse matrices of which the final size is not known initially. This was done by adjusting the shape property of an object inherited from the dok_matrix class. My problem is that in scipy 0.6 the dok_matrix is no longer re-sizable in this way. Is there a specific reason for this? Any suggestions will be welcomed. Thanks Evan -- visit http://randomestrandom.blogspot.com

1 0

Memory quirk
by Anand Patil 03 Nov '07

03 Nov '07

Hi all, The following script leaks memory when an array with dtype=object is used instead of a list. Is this a bug, or a behavior I should be expecting? Any help is much appreciated. Thanks, Anand from numpy import * import time import gc class MyObject(object): def __init__(self): self.data = zeros(100000, dtype=float) # This leaks memory... self.stufflist = array([self], dtype=object) # This doesn't # self.stufflist = [self] for i in xrange(10000): q=MyObject() time.sleep(.01) gc.collect()

2 1

Re: [SciPy-user] Memory quirk
by Anand Patil 03 Nov '07

03 Nov '07

Hi Travis, thanks for your help. I think I get it but your response to my first posting didn't come through, would you mind resending? Also, if possible it would be nice to have this information in Guide to NumPy. I had no idea numpy did its own garbage collecting, so I had no clue where to start looking. Cheers Anand ------------------------------ > > Message: 4 > Date: Sat, 03 Nov 2007 01:55:35 -0500 > From: "Travis E. Oliphant" <oliphant(a)enthought.com> > Subject: [SciPy-user] [Fwd: Re: Memory quirk] > To: SciPy Users List <scipy-user(a)scipy.org> > Message-ID: <472C1B67.1000703(a)enthought.com> > Content-Type: text/plain; charset="iso-8859-1" > >

1 0

communications toolbox for scientific python?
by Jose Unpingco 03 Nov '07

03 Nov '07

Is there anything like the Matlab communications toolbox in scientific python? Otherwise, is there anybody out there interested in putting one together? Thanks.

2 1