Попов Сергей Александрович
Decision Tree Ensembles on Tabular Data
Науки о данных
Nowadays, there are many machine learning problems on which deep neural networks (DNNs) outperform any other approaches by a notable margin, namely computer vision, natural language processing and speech. However, the main instrument on heterogenous tabular data is gradient boosting decision trees method. In this paper, we introduce new architecture that designed to work with tabular data with numeric features. Our experiments shows that proposed approach outperforms the competitors on most of the datasets.