手气是什么原因引起的

This project aim to group similar messages together, to reduce redundancy.
1. Find similarities between messages: I used minHash algorithm to convert each message to a vector of 100 hash values, by comparing the hash value vectors, I can get the difference between any two texts, hence predict the similarity. In order to guarantee accuracy, I also applied cosine similarity and Levenstein distance algorithm to the same dataset, and make sure the results are consistent .
2. Applied… 展开

This project aim to group similar messages together, to reduce redundancy.
1. Find similarities between messages: I used minHash algorithm to convert each message to a vector of 100 hash values, by comparing the hash value vectors, I can get the difference between any two texts, hence predict the similarity. In order to guarantee accuracy, I also applied cosine similarity and Levenstein distance algorithm to the same dataset, and make sure the results are consistent .
2. Applied unsupervised clustering algorithm including agglomerative method and Expectation Maximum.
The results of this project achieved high efficiency and accuracy. 收起

This project provides useful statistic data for Prime related messages of Amazon's website, the data includes daily average hits and daily average sign up. The purpose of this project is to better aid internal users for coming up with better marketing strategies, hence improving Prime's
1. Retrieve average daily hits data by sending http request to an internal monitoring website and parse http response. I used multi-threading to improve efficiency, the total retrieving time is reduced from… 展开

This project provides useful statistic data for Prime related messages of Amazon's website, the data includes daily average hits and daily average sign up. The purpose of this project is to better aid internal users for coming up with better marketing strategies, hence improving Prime's
1. Retrieve average daily hits data by sending http request to an internal monitoring website and parse http response. I used multi-threading to improve efficiency, the total retrieving time is reduced from 1 hour to 3 minutes.
2. Data Warehouse, to get daily sign up data associated with each message, I need to connect to data warehouse. I used JDBC for connection, and apply SQL query to retrieve required data.
3. User authentication: the data are credential, hence when displaying the results, they should be invisible to users outside of Prime organization. I used LDAP protocol to analysis users' role, and rendering different web pages dynamically according to the authentication results.
This project is highly appreciated by the whole team and prime market managers, it's is now widely used internally in prime organization. 收起

We are implementing an iphone's app, to detect smiling face in order to enhance the functionality of camera. To be specific, the camere with this detection function is able to eliminate the scenario where people close their eyes while taking the picture.

This is a research project of CCLS lab of Columbia. The main feature of this application is to read the scanned version of articles of New York Publication Library, using the implemented machine learning algorithm to accurately classify the newspapers, to be specific, assign labels like political, sports, ethic to categorize the papers. My job is to design and development part of the website, on which, users can see the scanned article, then they can manually label these papers, and submit the… 展开

This is a research project of CCLS lab of Columbia. The main feature of this application is to read the scanned version of articles of New York Publication Library, using the implemented machine learning algorithm to accurately classify the newspapers, to be specific, assign labels like political, sports, ethic to categorize the papers. My job is to design and development part of the website, on which, users can see the scanned article, then they can manually label these papers, and submit the result. By collecting this manual annotations, we can firstly use them as training data to train our machine learning algorithm, secondly, we can use them to compare to the result of our algorithm and refine our algorithm. 收起

This is a project for Software Engineering course. This is a web application, it support users to register for their own account, then login to see the current on sale items collected from Taobao(the amazon of China). We mainly used JSP on Spring+Hibernate to implement this.
Our project got high score which made us rank the top in class.

This application mainly support users of eBay to upload a video from their iphone or other smart phone for their items. Besides, users can add a new item, and also view informations about their existing item. My job is the development and design part of uploading video to Youtube, after getting the short link, embed it into the user's item's information.
Our project won the 2nd prize in the "Young Skunk Competition" of eBay. Our idea was highly appraised by VPs. And they decided to build up… 展开

This application mainly support users of eBay to upload a video from their iphone or other smart phone for their items. Besides, users can add a new item, and also view informations about their existing item. My job is the development and design part of uploading video to Youtube, after getting the short link, embed it into the user's item's information.
Our project won the 2nd prize in the "Young Skunk Competition" of eBay. Our idea was highly appraised by VPs. And they decided to build up a group to realize our idea, since after investigation, this new feature will truly attract more user, thus increasing total transaction and the revenue of company. 收起

We build a email analysis system using Hadoop, the purpose of our project is to find top 5 senders and receivers in the whole email system. The method used are email corpus, we process the data by MapReduce mechanism, and get the total sending mail numbers and receiving mail numbers for each user.

We build a yelp companion system, an iphone application. Using this application, users are able to find all the restaurant near their current location, for each restaurant we provide a recommendation list for the most preferred dishes.
For the server end, each time users request with their geo-location, we interact using yelp API, then get the nearby restaurant, after that, we extract all the comments for the restaurant, then process with NLP technique to find the most favorable dishes.

We are implementing a detecting system, where we use camera to track people's eye gestures, especially pupil movement detection and blink detection.
After, we implement two application prototypes to demonstrate the feasibility of our research. One is Eye Drawing, people can draw images on screen by moving eyes, and change color of drawing lines by blinking. The other one is Eye Mouse, which allows users to move mouse cursor by eyes.

This is a project of Cloud Computing course. We use AWS for language process, to be specific, we extract the comments for each restaurant from yelp, extract the recommended dishes and rendering the results to users.
This is an application on iphone.