Tabular representation learning
There is an intriguing question in neural networks.
Why deep neural networks does not surpass trees-based models until today?
Essa seria uma importante questão. Olha os modelos que se tem. ChatGPT, Claude. They demonstrate so powerful intelligence, but still, for this task, they can not solve with the rapidez e grande desempenho de modelos como XGBoost.
I have been considering this research line since 2024 with some ideas.
Today, my paper was accepted to ICLR Workshop.
Really special.