VladoUzunangelov/aklimate: AKLIMATE : Algorithm for Kernel Learning with Approximating Tree Ensembles

AKLIMATE uses a stacked learning approach with Random Forest(RF) base learners and a Multiple Kernel Learning (MKL) stacked learner. Each RF is trained on a collection of features drawn from heterogeneous data that correspond to a particular biological prior - a pathway, an experiment, or a computational prediction. The trees for an individual RF are then converted to an RF kernel that empirically captures the degree of similarity between training samples based on that RF's biological prior. The stacked MKL operates on the most predictive RF kernels to create an optimal meta-kernel that is used for the final predictions.

install.packages("remotes")
remotes::install_github("VladoUzunangelov/aklimate")