16th International Conference on Information Assurance and Security on the World Wide Web Dec 15-18, 2020

Dr. Ketan Kotecha, Head Symbiosis Centre for Applied Artificial Intelligence was invited as Speaker @ Mirlabs Conference in USA

Domain Adaption in Deep Learning - Theories and Applications

Deep learning has to create revolutions in the field of Artificial Intelligence. However, there is an unrealistic assumption that training and test data comes from the same distribution. This assumption does not hold in practical applications. An example includes, an autonomous car designed for Europe won't work directly in Indian scenario being visual data will be quite different. Spam email classifier intended for the general purpose may not work well for a specific user. Not only that there are applications where domain adaption will save tremendous efforts if implemented with the proper insight.   There are also attempts to generalise the algorithms for any domain leads to Domain Generalization(DG). Combining DA/DG to Few short learning and meta-learning also is a need of the day and a lot of work is in progress in those areas. 

The talk shall throw light on defining Domain Adaption and allied areas like transfer learning, few-shot learning, meta-learning and multitask learning. Also, few states of the algorithms in these areas, along with applications, will be discussed. The talk will end, giving a few directions for research.