Current Research Themes:

A. Smart Decision Making

We are developing a brand new algorithm framework for multi-agent decision making scenarios. We mainly exploit reinforcement learning (DQN, PPO), evolution strategies and transfer learning techniques to help agents conduct collaborative learning and make mission-critical decisions in complex environments. This is expected to be our flagship research theme supported by significant funding from the industry.

Recent Research Theme:

A. Massively Parallel Data Mining

We use the latest GPU computing architecture (CUDA 7.X and NVIDIA Kepler K80) to generate highly efficient (one to two orders of magnitude faster) and effective (more precise and robust) solutions for tackling large scale real-world data mining and optimization problems.

We are proud to host the NVIDIA GPU Education Center.  Award  Poster 1  Poster 2

Publications: JDSA  PAKDD  CIKM

B. Efficient Clustering Using Boundary Information

We investigate clustering techniques based on the boundary information of data. Instead of having all the raw data involved in the clustering process, only a small portion of data (5%-10%) is strategically extracted and fed into traditional clustering methods, which can dramatically improve the efficiency of clustering algorithms without compromising the quality of clustering.

Publications: Neurocomputing  PRL

C. Robust Clustering in Strong Noise Settings

We investigate clustering techniques capable of discovering complex clusters in datasets with a significant level of noise (80%), which is extremely challenging for traditional techniques. We employ density based multiscale analysis and various statistical techniques to reliably identify noisy objects from "clustered" objects.

Publications: AI 2017 (Award)

D. Vision Based Dynamic Hand Gesture Recognition for Touchless Human Computer Interaction

We develop robust vision based hand gesture recognition techniques using optical cameras, Kinect and Leap Motion, which are capable of reliably detecting, tracking and recognizing flexible human hand gestures under complex environments (unfavorable lighting conditions, occlusions and background noise) and have found exciting applications in our everyday life (smart TV, smart phone, virtual reality and computer games).

Research Showcase:  Poster

Other Research Activities:

Data Mining

• Classification of Imbalanced Datasets

• Ensemble Learning

• Sampling

• Social Computing & Recommendation

• Spatial Database

• Applications: CRM, Early Diagnosis of Alzheimer's Disease, Games

Evolutionary Computation

• Estimation of Distribution Algorithms

• Landscape Generator

• Racing Techniques

• Applications: MRI, Wireless Sensor Networks

A brief introduction of my previous research experience: here

I have been working with a number of nice people:

Alumni:

^* Award for Master's Thesis Excellence, Tsinghua University;
^# Award for Outstanding Master's Graduate, Tsinghua University/Beijing City