GANDALF: Gated Adaptive Network for Deep Automated Learning of Features
This addresses the need for high-performance deep learning models in tabular data applications, though it appears incremental as it builds on existing architectures with specific improvements.
The authors tackled the problem of developing an efficient and interpretable deep learning architecture for tabular data, resulting in GANDALF, which outperforms or matches state-of-the-art methods like XGBoost and FT-Transformers on multiple benchmarks.
We propose a novel high-performance, interpretable, and parameter \& computationally efficient deep learning architecture for tabular data, Gated Adaptive Network for Deep Automated Learning of Features (GANDALF). GANDALF relies on a new tabular processing unit with a gating mechanism and in-built feature selection called Gated Feature Learning Unit (GFLU) as a feature representation learning unit. We demonstrate that GANDALF outperforms or stays at-par with SOTA approaches like XGBoost, SAINT, FT-Transformers, etc. by experiments on multiple established public benchmarks. We have made available the code at github.com/manujosephv/pytorch_tabular under MIT License.