Kwang-Sung Jun

I'm 3rd year graduate student at Computer Science Department, University of Wisconsin-Madison. I obtained B.E. from School of Computing, Soongsil University in Republic of Korea (not North...). I majored in computer science and minored in mathematics. My research interests lie on statistical machine learning. I am currently working with Professor Xiaojin (Jerry) Zhu as an RA, and organized Non-IID Reading Group in Fall 2010.

Contact

email: ['deltakam' '@' 'cs' '.wisc' '.edu']

6397 Computer Science and Statistics
1210 W. Dayton St.
Madison, WI 53706

Courses Taken:

• CS766: Computer Vision (Zhang, Fall 2010)
• Course Project: Single View Modeling
• Implements Single View Model which recovers 3D scene from a single image. Interactive object segmentation is also implemented and Matlab GUI environment is built for user's convenience.
• CS576: Introduction to Bioinformatics (Dewey, Fall 2010)
• CS838: Computational Cognitive Science (Coen, Fall 2010)
• Course Project: Model Selection Using Rademacher Complexity
  • Identifies Rademacher complexity’s ability in model selection. Results over synthetic datasets show that model selection by upperbound of future error rate based on Rademacher complexity is inferior than k-fold cross validation in dataset of complex structure where highly expressive hypothesis space gives lower error rate.
• CS769: Advanced Natural Language Processing (Zhu, Spring 2010)
• Course Project: Detecting Fights in Video Comments
• Collected 400 conversations from YouTube comments with script language, and detected fight with 84% accuracy. Support Vector Machine and Naive Bayes were used to train and classify.
• CS760: Machine Learning (Shavlik, Spring 2010)
• Course Project: Implementing and Evaluating Dirichlet Process Mixture Model
• Implementation of Dirichlet Process Mixture model and evaluation, and identification of its predictive power in classification context. It shows inferior accuracy than baseline methods in real-world dataset while it was competitive for chosen synthetic dataset
• CS525: Linear Programming Methods (Wright, Fall 2009)
• CS731: Advanced Artificial Intelligence (Page, Fall 2009)
• Course Project: Implmenting and Analyzing Naive Bayes and Tree Augmented Naive Bayes
• Compares Naive Bayes classifier and Tree Augmented Naive Bayes on toy dataset. TAN works better only when there is enough data.

Research Interests:

• Statistical machine Learning
• Semi-Supervised Learning
• Online Learning