Projects

Switching from Agricultural Economics to Applied Mathematics, while always keeping my eyes on Computer Science, I've done several interesting projects during my undergraduate life.

Latest Projects

Snowcast Radio Station

A streaming multi-client supported terminal music station. Implemented both in C and Go.

[Documentation] [C Implementation] [Go Implementation]

Canonical Correlation Analysis in Image Classification and Annotation

Applied Mathematics double major graduation dissertation, School of Information

Implemented 3-veiw CCA model to employ I2I(Image to Image), T2I(Tag to Image) and I2T(Image to Tag) search

[Github]

Predicting the most popular New York Times blog(MIT Analytics Edge Course Project)

What makes online news articles popular? Newspapers and online news aggregators like Google News need to understand which news articles will be the most popular, so that they can prioritize the order in which stories appear. My ensembled glm+random forest model got ROC 0.90672 on test dataset and ranked 102/2923 on the final leadboard!

[Source Code ][Kaggle Profile]

Gomoku (a.k.a. Five Stones in a Row)

Interactive Gomoku Game between human player(yourself!) and computer AI. Take a try!

More platforms(Mobile Web, iOS and Android) will be supported soon.

[Project Page]

Web Proxy

Carnegie Mellon University 15-213 course project. This program runs as a proxy. It works by receiving HTTP requests from client, redirecting the requests to server and send the result back to client. Socket programming is used intensively.

[Source Code ]

Heap Allocator

Stanford CS 107 course project, grouped with Siyu Yang(Princeton). After the failure on Buddy's System approach, we successfully implemented a modified version based on implicit list design. Our heap allocator outperforms the full credit line by 32% and 15% in terms of utilization and throughput, respectively.

[Source Code]

Strongly Connected Components and their properties

Stanford CS 161 course project, grouped with Yifei Yang(USC). In this project we implemented an iterative version of Trajan's algorithm to find strongly connected components in directed graphs. In the report, properties such as degree of distribution, SCC diameters are further explored.

[Report] [C++] [Python]

Study of United States Local Food Market--a field study in Davis

Agricultural Economics and Management major dissertation, School of Agricultural Economics and Rural Development

Local food market is an innovative form of agricultural product market, which emphasizes a tighter geographical and social relationship between producers and consumers. This article introduces the development of local food market in United States and its corresponding supportive policies.

[PDF] [Slides]

Discovering the Best Coaches Ever--A Modern Approach Of Analytics Hierachy Process

Research paper written for Mathematical Competition of Modeling(MCM 2014), grouped with Shaoyang Wang(RUC) and Anqi Guo(RUC). Our research goal is to construct a coach evaluation model which could find the top 5 all time coaches among all possible sports. Specifically, we will first build our model based on college football, basketball and baseball coaches data. During this process, Satty's Analytics Hierachy Process will be used intensively with modifications. Next, we use some reliable online ranking sources to validate the correctness of output. By setting up an innovative algorithm, we can readily train our model until the evaluation function converges to its maximum. Afterward, we furthermore introduce non-quantitative factors into our model. Finally, we discuss the potential problems and those procedures that can be improved.

[PDF] [Source Code]

Image Filtering and Hybrid Images

Brown CSCI 1430 course project. The goal of this assignment is to write an image filtering function and use it to create hybrid images using a simplified version of the SIGGRAPH 2006 paper by Oliva, Torralba, and Schyns. Hybrid images are static images that change in interpretation as a function of the viewing distance. The basic idea is that high frequency tends to dominate perception when it is available, but, at a distance, only the low frequency (smooth) part of the signal can be seen. By blending the high frequency portion of one image with the low-frequency portion of another, you get a hybrid image that leads to different interpretations at different distances.

[Poster]

Regular Paper Fragment Stitching--Mathematical Model and Implementation

Research Paper written for China Mathematical Competition of Modeling(CMCM 2013), grouped with Jiaxuan Sun(RUC) and Ruqi Zhang(RUC). Fragment Stitching is the process of recovering document's original appearance. Our research paper proposes and experimented different ways of stitching rectangular subimages. Both cosine distance andL-norm regularization methods are applied. We found that cosine distance metrics fit well with this particular problem. When the edge length is too small to retrieve enough information, manual intervention will be introduced in order to guarantee a high accuracy.

Award: Second Prize

[PDF] [Source Code]