GATE: Gated Additive Tree Ensemble for Tabular Classification and Regression

We propose a novel high-performance, parameter and computationally efficient deep learning architecture for tabular data, Gated Additive Tree Ensemble(GATE). GATE uses a gating mechanism, inspired from GRU, as a feature representation learning unit with an in-built feature selection mechanism. We combine it with an ensemble of differentiable, non-linear decision trees, re-weighted with simple self-attention to predict our desired output. We demonstrate that GATE is a competitive alternative to SOTA approaches like GBDTs, NODE, FT Transformers, etc. by experiments on several public datasets (both classification and regression). The code will be uploaded as soon as the paper comes out of review.