Welcome to Data Mining Laboratory in Department of Computer Science at KAIST. The main theme of our lab is Large-scale Data Mining, and research interests include but are not limited to devising scalable algorithms for big graph (tensor) mining, and developing large-scale graph mining platforms.
The production of research for multi-dimensional big data analysis system receives wide media coverage (YONHAP NEWS Agency). The system analyzes 10 to 100 times larger data compared to existing system by efficiently designed distributed algorithm. The result of research will be presented to ICDE 2015, a top tier database conference.
Undergraduate student Mr. Kijung Shin won the HumanTech award (gold, 1st in Computer Science) from Samsung, from his paper "BEAR: Block Elimination Approach for Random Walk with Restart on Large Graphs". Ms. student Mr. Ho Lee also won the HumanTech award (honorable mention, 6th in Computer Science area), from his paper "Fast Graph Mining with HBase". HumanTech award is given to best papers in the engineering area, and it is the most prestigious award among such kind in Korea. Congratulations!
The paper "HaTen2: Billion-scale Tensor Decompositions" is accepted to ICDE 2015, a top tier database conference. The paper is on scaling up two important tensor decomposition methods, Tucker and PARAFAC, on MapReduce.
A research track paper and a demo paper are accepted to ICDM 2014, a top tier data mining conference. The research track paper "Distributed Methods for High-dimensional and Large-scale Tensor Factorization" proposed efficient distributed algorithms for large scale tensor factorization which can be used for recommendation. The demo track paper "Eventera: Real-time Event Recommendation System from Massive Heterogeneous Online Media" demonstrates a real-time event recommendation system.
3 regular papers are accepted to CIKM 2014, a top tier data mining conference.
Ph.D. student Ha-Myung Park won the HumanTech award (bronze) from Samsung. His paper is titled "An Efficient MapReduce Algorithm for Triangulation in a Very Large Graph" and it proposed an efficient MapReduce algorithm for finding triangles in large graphs.
Ph.D. student Yongsub Lim won the grant award from the "Venture Research Program for Graduate and PhD Students" from KAIST, for his proposal on distributed real time graph stream mining. The grant is highly selective (10 students in KAIST this year), and aimed for providing active support for creative and influential ideas that entail high risk.
MS student Inah Jeon won the 2013 Qualcomm Innovation Award. The award is a fellowship given to high quality papers. Her paper is titled "GigaTensor: Scaling Tensor Analysis Up By 100 Times - Algorithms and Discoveries" and it focused on the tensor analysis for finding patterns and anomalies in billion-scale real-world tensor using Hadoop.
Prof. U Kang won Honorable Mention for the 2013 KDD Dissertation Award, which is the highest honor for a data mining thesis. His dissertation is titled "Mining Tera-Scale Graphs: Theory, Engineering and Discoveries" and it focused on the award winning PEGASUS system for finding patterns and anomalies in billion-node graphs using thousands of machines. U received a certificate of recognition during the opening ceremonies at the upcoming KDD Conference in Chicago.