genRBF-missing

This package implements genRBF (see Generalized RBF kernel for incomplete data). The code is written in Python and Cython.

Requirements:

  • numpy (1.12.1 or higher),
  • cython (0.25.2 or higher),
  • gcc, g++ (5.4.0 or higher).

To estimate a Gaussian density from incomplete data used in genRBF, we applied R package norm.

Installation

Go to directory genRBF-missing/genRBF_source/ and run the following instruction in terminal:

./build.sh

or

python setup.py build_ext --inplace

Usage

The file 'main_demo.py' applies genRBF to SVM classifier. To run this demo, type the following command in terminal:

./main_demo.py ./data/

A directory genRBF-missing/data/ contains exemplary data with missing attributes: train data, test data and their labels. Files mu.txt, cov.txt contain covariance matrix and mean estimated over train data and were created with use of norm.R:

./norm.R ./data/train_data.txt ./data/