Oxytrees: Model Trees for Bipartite Learning

TL;DR

Oxytrees are a novel model tree approach for bipartite learning that significantly reduces training time while maintaining or improving predictive accuracy across diverse datasets.

Contribution

This paper introduces Oxytrees, a scalable, proxy-based biclustering model tree with a new leaf-assignment algorithm and linear models, enhancing efficiency and generalization in bipartite learning.

Findings

Up to 30-fold faster training compared to state-of-the-art methods.

Competitive or superior predictive performance on 15 datasets.

Effective in inductive learning scenarios.

Abstract

Bipartite learning is a machine learning task that aims to predict interactions between pairs of instances. It has been applied to various domains, including drug-target interactions, RNA-disease associations, and regulatory network inference. Despite being widely investigated, current methods still present drawbacks, as they are often designed for a specific application and thus do not generalize to other problems or present scalability issues. To address these challenges, we propose Oxytrees: proxy-based biclustering model trees. Oxytrees compress the interaction matrix into row- and column-wise proxy matrices, significantly reducing training time without compromising predictive performance. We also propose a new leaf-assignment algorithm that significantly reduces the time taken for prediction. Finally, Oxytrees employ linear models using the Kronecker product kernel in their leaves, resulting in shallower trees and thus even faster training. Using 15 datasets, we compared the predictive performance of ensembles of Oxytrees with that of the current state-of-the-art. We achieved up to 30-fold improvement in training times compared to state-of-the-art biclustering forests, while demonstrating competitive or superior performance in most evaluation settings, particularly in the inductive setting. Finally, we provide an intuitive Python API to access all datasets, methods and evaluation measures used in this work, thus enabling reproducible research in this field.