cppyml

Optimised Python extension for machine learning based on the ML++ C++ library.

© 2020 Roman Werpachowski.

Contents

• ordinary least squares (OLS) regression
• recursive OLS regression
• logistic regression
• classification and regression decision trees with pruning
• Gaussian E-M clustering and K-means

Documentation

Automatically generated documentation is here

Installation

Linux

Build from source using SCons and install using sudo python setup.py install:

scons mode=release
cd cppyml
sudo python setup.py install

Windows

1. Add a path to the libs directory of a Python distribution to the Visual Studio properties file LocalDependencies.props in the “Linker -> General -> Additional Library Directories” section. For example, I use C:\Users\Roman\anaconda3\libs. These paths determine the Python version required by the built module.

2. Build the solution in ReleaseStatic mode.

3. From the cppyml directory, run python setup.py install to install the Python module linked to ReleaseStatic build of the ML++ library. Remember to install the cppyml package in a Python environment with the same Python version as the one used to build it!

Example

import numpy as np
from cppyml import linear_regression

n = 25
x = np.random.randn(n)
y = 0.1 * x - 0.9 + 0.2 * np.random.randn(n)
result = linear_regression.univariate(x, y)

See also unit tests in tests.