ICML 2013 experiments for "Consistency of Online Random Forests" using rftk. To run the experiments you must clone and build rftk and rftk-colrf-icml2013. The rftk project has undergone significant refactoring since the paper was published. We recommend using the latest implementation but we also provide the commits required to reproduce the exact results from the paper.
Clone both projects
git clone https://github.com/david-matheson/rftk.git
git clone https://github.com/david-matheson/rftk-colrf-icml2013.git
If you wish to use the exact setup as the paper checkout the colrf-icml2013-camera-ready tag.
cd /path/to/rftk
git checkout -b colrf-icml2013-camera-ready origin/colrf-icml2013-camera-ready
cd /path/to/rftk-colrf-icml2013
git checkout -b colrf-icml2013-camera-ready origin/colrf-icml2013-camera-ready
However, we recommend using the latest implementation in the colrf-icml2013 branch.
cd /path/to/rftk
git checkout -b colrf-icml2013 origin/colrf-icml2013
cd /path/to/rftk-colrf-icml2013
git checkout -b colrf-icml2013 origin/colrf-icml2013
Build rftk
cd /path/to/rftk
scons
Add rftk to your PYTHONPATH
PYTHONPATH=/path/to/rftk/
export PYTHONPATH
Download and unzip data
cd /path/to/rftk-colrf-icml2013/source_data
wget http://www.cs.ubc.ca/~mdenil/projects/random_forests/usps_mog.tar.bz2
tar -jxvf usps_mog.tar.bz2
wget http://www.cs.ubc.ca/~mdenil/projects/random_forests/kinect_data.zip
unzip kinect_data.zip
cd /path/to/rftk-colrf-icml2013/
./run_advantage_of_forest.sh
./run_compare_to_offline.sh
./run_kinect.sh