Using Machine Learning methods to analysis students' video watching behaviors and learning behaviors on 10 Coursea courses. Clustering students by their learning behaviors then building a visualization interface for them to get information from other students with similar behaviors.
In this project we implemented a model that utilizes review text to improve the prediction for Amazon movie scores. Firstly we use NLP techniques to preprocess raw review text and build the corpus. Next we extracted topic distributions from the corpus using LDA, and built a Hidden Factor Model to learn and predict scores for products. To solve the cold-start problem, we regularize the traditional Hidden Factor Model with an corpus likelihood term and get a better result on scores predicting.
Analyse the data generated by a course on MOOCs in Tsinghua University. Using data mining and visualization techniques to classify and predict students' behaviours. A corresponding system has been developed and now is work for the course.
The project is to control a 6DOF robot arm to hold a table tennis stable on a table tennis racket. My job in the project including problem modelling, computer vision programming, control algorithm design.
see video at : https://www.youtube.com/watch?v=MugBPurFE54
Proficiency with Java/Python/C/C++/Matlab/R. Comfortable with developing on Linux.
Familiar with basic ML methods/models and their implementations in large scale version. Have experience with Map-Reduce and Hadoop.
Have experience on implementing search engine, collaborative filtering, social network analysis. Familiar with basic IR models (BM25, Indri, LeToRank).
Have knowdlege of both statistical and traditional NLP theory. Familiar with different parsing algorithm (CKY, LR, Earley, Agenda based, dependency paring). Have experience with NLTK.
Have knowdlege of Computer Vision and Digital Image Processing. Have experience on multimedia feature extraction (ASR, MFCC, CNN, SIFT) and multimedia data mining.
HTML/Django/D3.js
My job is mainly doing software supports for the design of a new intelligent electricity power meter. Works including embedded programming, real time OS migration and algorithm analysis.