Reinforcement learning and its applications to control and multi-agent systems is where Dr. Jing Peng's doctoral studies originated. In general terms, reinforcement learning is concerned with computational approaches to learning from reward. Here he considers how a learning agent can learn as quickly as possible from limited interaction with other agents and its environment.

Image retrieval, in particular content-based retrieval where he considers indexing schemes that allow flexible retrieval metrics to be created on the fly so that very large databases can be accessed efficiently and accurately.

Classification and data mining, is where he is particularly interested in adaptive metric nearest-neighbor techniques and compact subspace representation for building robust classifiers from limited training data.

Classification and Data Mining

Adaptive metric nearest-neighbor techniques and compact subspace representation for building robust classifiers from limited training data.