Extreme Multi-labelling with Neural Networks

Abstract

Extreme multi-label learning (XML) or extreme multi-label classification is no longer a theoretical problem due to the enormous growth of data. Those commonly used technologies each as web page tagging (web labelling) and product recommendation are typical example of using the techniques of extreme multi-labelling. Extreme multi-label learning aims to correctly tag a data point with the most relevant class labels by machine from an enormous set of class labels by training a classifier (Bhatia, Jain, Kar, Varma, & Jain, 2015). In this project, extensive experiments on benchmark datasets from "The Extreme Classification Repository" will be conducted. I aim to improve the currently existing extreme multi-labelling learning algorithm with using different deep neural network architectures. The result is satisfactory.