Home / experiments
Name Modified Size InfoDownloads / Week
Parent folder
data_arrays 2016-07-07
results 2016-07-07
utils_sampling.pyc 2016-07-07 1.6 kB
utils_data.pyc 2016-07-07 7.6 kB
utils_explanations.py 2016-07-07 1.0 kB
utils_sampling.py 2016-07-07 1.3 kB
utils_classifiers.py 2016-07-07 6.6 kB
utils_classifiers.pyc 2016-07-07 6.2 kB
utils_data.py 2016-07-07 8.5 kB
run_experiment.py 2016-07-07 4.5 kB
prediction_difference_analysis.py 2016-07-07 14.2 kB
prediction_difference_analysis.pyc 2016-07-07 9.7 kB
data_converter.py 2016-07-07 8.7 kB
Totals: 13 Items   69.9 kB 0
This repository holds the implementation for the paper "Explaining individual classifier decisions for microbiota diagnosis", by A. Eck, L.M. Zintgraf, E.F.J. de Groot, T.G.J. de Meij, T.S. Cohen, P.H.M. Savelkoul, M. Welling, and A.E. Budding.

The important class is in prediction_difference_analysis.py, which handles all the computations. For a given input vector to a probabilistic classifier, an according relevance vector can be calculated by calling univ_rel().

The script run_experiment.py can be executed to run a simple experiment, returning for each sample in the dataset a relevance vector. Settings can be adjusted in the script (like which classifier to use).
Source: README.txt, updated 2016-07-07