Acknowledgments

We would like to express our sincere gratitude to the following repositories and projects for providing helpful components, functions, and ideas that have contributed to the development of TALENT. These projects have significantly supported the integration of various methods into our toolbox:

  • RTDL-revisiting-models : Provided valuable components for rethinking tabular data models.

  • RTDL-num-embeddings : Provided key techniques for numerical embeddings in tabular data.

  • Tabular-dl-tabr : Offered insights into deep learning models for tabular data.

  • DANet : Contributed to the development of correlated feature grouping for tabular data.

  • TabCaps : Provided the framework for capsule networks used in tabular data tasks.

  • DNNR : Enabled enhanced KNN-based methods for tabular data predictions.

  • PTaRL : Supported the creation of a prototype-based regularization framework.

  • Saint : Provided row and column attention mechanisms for tabular data modeling.

  • SwitchTab : Helped in the implementation of self-supervised methods for tabular data.

  • TabNet : Provided attention-based feature selection techniques.

  • TabPFN : Offered a generalizable pre-trained network for tabular tasks.

  • Tabtransformer-pytorch : Contributed to transforming categorical features into contextual embeddings.

  • TANGOS : Provided key insights into regularization for tabular learning models.

  • GrowNet : Helped develop gradient boosting frameworks for shallow networks.

  • HyperFast : Supported the use of meta-trained hypernetworks for tabular tasks.

  • BiSHop : Provided key elements of sparse Hopfield models for tabular learning.

  • ProtoGate : Supported the development of prototype-based models for high-dimensional data.

  • Pytabkit : Offered helpful tools for efficient tabular data handling.

  • Excelformer : Contributed techniques for attention-based tabular models.

  • GRANDE : Provided insights for tree-mimic models using gradient descent.

  • AMFormer : Supported transformer-based methods for tabular feature interactions.

Additionally, we would like to acknowledge the authors of these works for their invaluable contributions to the field of machine learning on tabular data. Without their innovations, TALENT would not have reached its current state.

If you find any components or resources in TALENT useful in your research, we encourage you to cite these repositories and papers alongside TALENT.