[SciPy-Dev] Expanding Scipy's KDE functionality

Daniel Smith smith.daniel.br@gmail....
Wed Feb 6 16:48:13 CST 2013


I've updated the github repository with some new code:


I added a data generating process class and subclassed it with the
examples from the Botev, Grotowski, and Kroese paper. I could have
subclassed scipy's rv_continuous, but I thought that class was a lot
bigger and bulkier than we need. The script test_plots.py produces a
bunch of pdf's that show how well the KDE estimate aligns with the
probability distribution function. Next, I will write some code to
compare the square error of the new method with the square error from
SciPy's current implementation.

Finally, I am still planning on writing some Cython code to calculate
the derivative using a minimal number of Fourier basis vectors.


More information about the SciPy-Dev mailing list