With the increasing popularity of applying machine learning for various artificial intelligence (AI) problems, a plethora of machine learning techniques are being developed. However, a critical limitation of conventional machine learning paradigms arises from the scarcity and expense of (labelled) data as popular deep learning methods often have millions of parameters and require thousands (if not millions) of samples to train an effective model. In this aspect, machine intelligence is considered inferior to human intelligence, as humans have the ability to learn rapidly from very limited data. This project explores new approaches to investigate the fundamental research problem of learning from small (labelled) data, called one-shot learning or few-shot learning, and will develop new algorithms and techniques to devise one-shot learning machines with human-like learning capabilities. The overall objective of this project is advance AI research – both in bringing AI closer to human intelligence and solving real-world problems.
Research Technical Area
- Machine learning
Benefits to the society
The research on one-shot learning unlocks potential opportunities for further application in areas such as visual image recognition, healthcare analytics and cybersecurity intelligence etc. In addition to building capability and developing AI fundamentals, this development is also an enabling factor as Singapore pushes for the Smart Nation initiative.
- Chenghao Liu, Xin Wang, Tao Lu, Wenwu Zhu, Jianling Sun, Steven C. H. Hoi:
Discrete Social Recommendation. AAAI 2019: 208-215
- Chen Zhang, Steven C. H. Hoi:
Partially Observable Multi-Sensor Sequential Change Detection: A Combinatorial Multi-Armed Bandit Approach. AAAI 2019: 5733-5740
- Yang Li, Jianke Zhu, Steven C. H. Hoi, Wenjie Song, Zhefeng Wang, Hantang Liu:
Robust Estimation of Similarity Transformation for Visual Object Tracking. AAAI 2019: 8666-867
- Chenghao Liu, Tao Lu, Xin Wang, Zhiyong Chenghao, Jianling Sun, Steven C. H. Hoi:
Compositional Coding for Collaborative Filtering. SIGIR 2019: 145-154