jerkern / pyparticleest Goto Github PK

Hi!

Thanks for a very interesting project. I downloaded and installed it to play around with, and got a fatal warning when running setup.py:

fatal error: 'numpy/arrayobject.h' file not found 

This makes the compiled C speedup unavailable. If other users encounter this it can be fixed by adding import numpy and include_dirs=[numpy.get_include()] to the extensions import in setup.py, like this:

import numpy
...

if (not on_rtd):
extensions = [Extension("pyparticleest/utils/ckalman", ["pyparticleest/utils/ckalman" + ext], include_dirs=[numpy.get_include()]),
              Extension("pyparticleest/utils/cmlnlg_compute", ["pyparticleest/utils/cmlnlg_compute" + ext], include_dirs=[numpy.get_include()])]

Hi, sorry this is more of a query than an issue. I am hoping to use Standard Nonlinear Model for filtering/smoothing 2d indoor positional coordinates. I have the corresponding blueprint map as well. Would modelling x & y (coordinates) separately make sense as current implementation seem to work-with/simulate only 1d data? also not sure how to include landmark info at each step. Many thanks!
P.S. thanks for your wonderful work and amazing explanation.

I was trying to figure out where your expression for kalman.lognormpdf came from, so I went back to the functional expression

and took the ln of that to find

I understand how this is the same as your code
-0.5 * (S.shape[0] * l2pi + np.linalg.slogdet(S)[1] + np.linalg.solve(S, tmp).T.dot(tmp))
but if my function f is correct, then your expression is missing that summation np.sum(tmp)

Thoughts?

-ben

When attempting to import this library for Python3.5, I get the infamous "No module named 'exceptions'" error.

It really would be useful to see this library updated for Python3.

Hello,

I set an instance of StdNonLin model, with two fixed parameters, and I was able to use the simulator.Simulator class to do filtering and smoothing successfuly. Now I want to do parameter estimation for these two parameters. Do you by any chance have any examples for that?
I could not get the source code for the paramest class and also encountered the error module 'pyparticleest.paramest' has no attribute 'paramest' when I tried to run

import pyparticleest.paramest as paramest

est = paramest.paramest.ParamEstimation(model, u=None, y=obs)

as in the API example.

Thanks for your cool package, looking forward to use it for parameter estimation too.