ngmix

Gaussian mixture models and other code for working with for 2d images, implemented in python. The code is made fast using the numba package. Note the old c-extension based code is still available in the tag v0.9.5

For some examples, please see the wiki.

dependencies

• numpy
• numba

optional dependencies

• scipy: optional needed for image fitting using the Levenberg-Marquardt fitter
• galsim: optional for doing metacalibration operations.
• skikit-learn: for sampling multivariate PDFs
• statsmodels: optional for importance sampling (multivariate student T distribution)
• emcee: optional for doing MCMC fitting: http://dan.iel.fm/emcee/current/ Affine invariant MCMC sampler.

installation

python setup.py install

installing numba

# by far the easiest way is using anaconda.
conda install numba

TODO

Make numba optional. There are many useful data structures in ngmix that do not require numba, it would be nice to make them available without the numba dependency. To implement this would be straightforward: wrap some of the imports in try/except blocks