KPDF - 0.2
Kernel-based Probability Density
The module contains functions for univariate and
density function estimation using a kernel-based approach (Silverman,
There are some other utility functions which provide
estimations of the optimal bandwidth parameter.
- Univariate estimators
- Multivariate estimators (with and without
along each axis). Dimension lower or equal to three.
- Multivariate Gaussian
It is entirely written in C to allow a fast
mainly for two and three dimensional estimations with several points.
To be able to use it, you will need:
If you want to be able to run the tests for
PDFs, you will also need:
- Python ;-)
- netCDF library 3.4 or later
- Scientific Python, by Konrad Hinsen
For UNIX platforms (the OLD way):
should suffice. The last step will probably require
- make -f Makefile.pre.in boot
- make install
Dec-2002. Michiel Jan Laurens
de Hoon, University of Tokyo,
Institute of Medical Science, has contributed a
binary Windows port of
the package and a setup.py script for the module and
source distribution, which can
now be compiled by the new method "python setup.py install".
These files DO NOT hold additional testing files, so, python test.py
will NOT work unless you download the old distribution. For Windows,
testing of high-dimensional PDFs will not be posible unless you
also install Scientific.IO.NetCDF at your own risk (it will not be
easy, though). The fact that the test do not work does not mean that
KPDF does not work. It simply means that the regression tests can not
proceed due to the lack of access to netCDF files in this platform.
Feb-2004. Michiel de Hoon,
University of Tokyo, Institute of Medical Science, Human Genome Center (http://bonsai.ims.u-tokyo.ac.jp/~mdehoon)
has contributed new installers for windows and a bugfix for the C
module, so that it currently checks that an array is being passed to
the code, instead of any other type of Python container. We have called
this the 0.2 version of the code and here are the source distribution
and the windows installers of version 0.2, kindly provided by Michiel.
The source distribution contains the files for the tests:
A README file in the distribution provides information
the parameters to the functions. For mathematical details, refer to the
original literature (Silverman, 1986). A thorough documentation in
is under preparation.
Europe, source tarball including a setup.py script but not tests.
USA, source tarball including a setup.py script but not tests
Europe (old source tarballs)
USA (old source tarballs)
Europe, Windows installer but no tests.
USA, Windows installer but no tests.
VERSION 0.2 (see above):
Any feedback from the users of the module will be
by the author.
It is released under the GNU Public License.
- 20000923 - Version 0.1: Added functions
create "linear" two and
three dimensional grids from NumPy arrays defining a set of coordinates
along each of the two or three dimensional spaces in a single call.
avoids the need to use complicated calls to map(), concatenate() and
plus the definition of some lambda expressions to use inside this
Rest of code untouched. See documentation for details.
- 20040223 -
Version 0.2: New installers for Python 2.1-Python 2.3. A new
check added to the C code to verify that it is receiving proper Numeric
arrays. Bug fix provided by Michiel de Hoon, University
of Tokyo, Institute of Medical Science, Human Genome Center.
B. W. Silverman (1986) Density Estimation for
and Data Analysis, 1st edition, Chapman and Hall, London, 175 pages.