XBNet : An Extremely Boosted Neural Network

  • 2021-06-09 17:31:50
  • Tushar Sarkar
  • 3

Abstract

Neural networks have proved to be very robust at processing unstructured datalike images, text, videos, and audio. However, it has been observed that theirperformance is not up to the mark in tabular data; hence tree-based models arepreferred in such scenarios. A popular model for tabular data is boosted trees,a highly efficacious and extensively used machine learning method, and it alsoprovides good interpretability compared to neural networks. In this paper, wedescribe a novel architecture XBNet, which tries to combine tree-based modelswith that of neural networks to create a robust architecture trained by using anovel optimization technique, Boosted Gradient Descent for Tabular Data whichincreases its interpretability and performance.

 

Quick Read (beta)

loading the full paper ...