Skip to content

zygmuntz/metric-learning-for-regression

Repository files navigation

Metric Learning for Kernel Regression

See http://fastml.com/good-representations-distance-metric-learning-and-supervised-dimensionality-reduction for description.

Download and install MLKR (Matlab; it doesn't seem to work in Octave).

Edit paths in the following scripts and run them:

mlkr_rescale_separately.m
rf_mlkr_rescaled_separately.r

mlkr_rescale_together.m
rf_mlkr_rescaled_together.r

The first script in each pair transforms the dataset, the second trains a random forest. The rescaling method doesn't seem to matter with this dataset, meaning you can just run one pair.

rf.py is Python code for random forest and pca_rf.py is the same thing with PCA thrown in, for comparison.