Bridging domains using world wide knowledge for transfer learning

ABSTRACT:

A major downside of classification learning is the dearth of ground-truth labeled knowledge. It is usually expensive to label new information instances for coaching a model. To solve this problem, domain adaptation in transfer learning has been proposed to classify target domain information by using another source domain information, even when the info may have completely different distributions. However, domain adaptation could not work well when the variations between the source and target domains are massive. In this paper, we have a tendency to design a unique transfer learning approach, referred to as BIG (Bridging Info Gap), to effectively extract useful data during a worldwide knowledge base, which is then used to link the supply and target domains for improving the classification performance. BIG works when the supply and target domains share the same feature space however totally different underlying knowledge distributions. Using the auxiliary source data, we tend to will extract a ¿bridge¿ that permits cross-domain text classification problems to be solved using standard semisupervised learning algorithms. A major contribution of our work is that with BIG, a large amount of worldwide data will be simply custom-made and used for learning in the target domain. We conduct experiments on several real-world cross-domain text classification tasks and demonstrate that our proposed approach will outperform many existing domain adaptation approaches considerably.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here