This repository contains resources for selecting statistically significant features using multiscale bootstrap. The test corrests for selection bias. The algorithm is described in our paper,
Lim, J., Yamada, M., Jitkrittum, W., Terada, Y., Matsui, S., Shimodaira, H.
More Powerful Selective Kernel Tests for Feature Selection
AISTATS 2020
Requires numpy
, matplotlib
, SciPy
, sklearn
. The package
can be installed with the following command
pip install git+https://github.com/jenninglim/multiscale-features
Once installed, you should be able to do import mskernel
without any error.
See notebooks
.
See experiments
for experiment setup and its corresponding
figures can be seen in figures
.