Projects related to machine learning & data mining
KDD Cup 2013: Author-Paper Identification
The KDD Cup 2013 started! This project is advised by Prof. Guocan Feng and supported by Guangdong Province Key Laboratory of Computational Science .
B.S. Thesis: Compressed Sensing and Related Learning Problems
Compressed sensing is among the emerging studies in computational science. Machine learning people also focus on this topic and are interested in the learning from sparsity. This thesis provided an overview of recent studies in CS and related ML topics, incl. math theories and proofs, fast BP algorithms, as well as the dictionary learning. My thesis supervisor is Prof. Haizhang Zhang.
'SYSU Street View' - Panorama of Yixian Road in SYSU
Sun Yat-Sen University, one of the 'Ten Univ. of Best Sight Campus', has a forest-like campus. We enjoy taking a deep breath when walking along the Yixian Road aligned with trees. We're proud of our beautiful campus and glad to share our feeling with you. So my friend who majored in Designing and I took photos of the Yixian Road manually, and work hard for auto image stitching to develop street-view-like panorama.
KDD Cup 2012 Track 1: A Hybrid Recommender System of Tencent Microblog
Precise recommendations in social networks present opportunities for novel machine learning. Our work described a recommender system of micro blog, which mined topics from messages of users by applying association rules and by combining collaborative and content-based filtering to discover their preferences. We selected and rated items according to the interests of users, as reflected by the topics in the messages. The results indicated a strong relationship between user interests and their acceptance on a given item. Several limitations of our system have led to future research. The revision of topic extraction is related to NLP studies. The analysis of sequential data, data pre-processing, and advanced item pre-selection are other probable extensions.
Predicting Diameters of Electrospinning Products
Electrospinning uses an electrical charge to draw very fine (typically on the micro or nano scale) fibers/balls/mixed products from a liquid . This work focused on simulating ambient influence on the size of electro-spinning products. We collected the data on product diameters as well as on molecular weight, solution concentration, electric potential, and on the distance between the capillary and the collection screen. By directly applying the RBF network model to the raw data that result in inaccuracy, we performed data pre-processing by fuzzy c-means clustering.
Citi Cup 2011: Bank Customers Rating System Based On Expert System and ANN
The precise rating of customers has a decisive impact on loan business. We constructed the BP network, which requires financial statements as the input and predicted ratings that range from A to AAA. To make flexible prediction, we developed another fuzzy expert system, which requested the analyst to answer raised questions. Finally, we combined these two models and provided a final rating by trade-off.
(To avoid legal problems, I omitted the name of the companies when packing the tar.gz file.)
A Brief Talk about the Relationship between Python and Linux [pdf] (in Chinese)
Fraction and Chaos [pdf] (in Chinese)
2012 MCM - Designing the Optimal Schedule of the White Water Rafting [pdf]
Hanoi - developed by the assembly language [demo]
Yingzhen Li and Ye Zhang. Generating Ordered List of Recommended Items: a Hybrid Recommender System of Microblog. 21st Machine Learning Summer School - Kyoto University, Sept 2012. [poster]
Yingzhen Li and Ye Zhang. Generating Ordered List of Recommended Items: a Hybrid Recommender System of Microblog. 20st Machine Learning Summer School - University of California, Santa Cruz, July 2012. [poster]
Naiben Huang, Yingzhen Li and Haobo Xia. Student Seminar of Neutrino Science and Time Travel, Sun Yat-sen University, Oct 2011 [abstract]