Edit on GitHub

Python wrapper

The wrapper provides an interface through which the methods of CherenkovDeconvolution.jl are exposed to Python. This short guide assumes you have read the Manual section.

Installation

  • Step 1: You need a working Julia installation on your system. To install Julia, please refer to the official instructions.
  • Step 2: The wrapper, including all dependencies, can then be installed via pip:
pip install https://github.com/mirkobunse/CherenkovDeconvolution.jl/archive/py.tar.gz

Usage

Python translations of the examples from the Manual section would be:

import CherenkovDeconvolution_jl as cd
from sklearn.naive_bayes import GaussianNB

# DSEA+ example
dsea = cd.DSEA(GaussianNB())
f_dsea = cd.deconvolve(dsea, X_obs, X_trn, y_trn)

# classical deconvolution example
run = cd.RUN(cd.TreeBinning(10)) # use up to 10 clusters
f_run = cd.deconvolve(run, X_obs, X_trn, y_trn)

# adaptive stepsizes in DSEA+
stepsize = cd.RunStepsize(cd.TreeBinning(10), decay=True)
dsea_plus = cd.DSEA(GaussianNB(), K=100, epsilon=1e-6, stepsize=stepsize)
f_dsea_plus = cd.deconvolve(dsea_plus, X_obs, X_trn, y_trn)

How it works

This wrapper is realized with PyJulia.

The pip installer takes care of preparing the existing Julia installation for the wrapper: it installs the Python bridge PyCall.jl, the wrapped package CherenkovDeconvolution.jl, and all of their Julia package dependencies.

Calling import CherenkovDeconvolution_jl in Python starts a Julia process in the background, which executes all Julia methods with the argument values provided by Python and returns the results.

Troubleshooting

The PyJulia bridge documents solutions to some of its frequent issues. Additional bug fixes are discussed in the issues of PyJulia.

Some of these solutions are already part of the installation process, see setup.py and CherenkovDeconvolution_jl/init.py in the py branch of this repository. If you encounter any other problems, please file an issue at our package.