MKLpy 0.3

A package for Multiple Kernel Learning scikit-compliant

MKLpy is a framework for Multiple Kernel Learning and kernel machines scikit-compliant.

This package contains:

• MKL algorithms * EasyMKL * Average of kernels * Soon available: GRAM, MEMO, SimpleMKL

• tools to operate on kernels, such as normalization, centering, summation, mean…;

• metrics, such as kernel_alignment, radius, margin, spectral ratio…;

• kernel functions, such as homogeneous polynomial and boolean kernels (disjunctive, conjunctive, DNF, CNF).

The ‘examples’ folder contains useful snippets of code.

For more informations about classification, kernels and predictors visit Link scikit-learn

requirements

To work properly, MKLpy requires:

• numpy

• scikit-learn (v. 0.20.0)

• cvxopt