Worked in a group to benchmark xRFM, a recent algorithm that fuses feature-learning kernel machines with an adaptive tree structure, against established methods including XGBoost and Random Forest across a range of tabular datasets. Analysed the strengths and limitations of xRFM in terms of predictive performance, training efficiency, and feature interpretability via the Average Gradient Outer Product (AGOP).
For the python code used to fit and evaluate all of the models click here: Link to GitHub