What I particularly like about this conference is the fact that the breakout sessions always include a tutorial session, which means that you can either choose to deepen your knowledge in a specific field of data mining by attending a regular sessions (e.g. on classification, clustering or special topics such as social networks) or you could broaden your horizon by attending a tutorial on a topic you know relatively little about. I decided to attend the Classification I session. The first talk was by Wei Bi (see picture below) titled “Hierarchical Multilabel Classification with Minimum Bayes Risk” (co-author: James T. Kwok). Hierarchical multilabel classification (HMC) allows an instance to have multiple labels residing in a hierarchy. Usually, H-loss is used, which penalizes only the first classification mistake along each prediction path. The authors propose a hierarchy-aware loss function that is more appropriate for HMC.

Next, the paper “Multi-Task Semi-Supervised Semantic Feature Learning for Classification“ was presented (authors: Changying Du, Fuzhen Zhuang, Qing He, and Zhongzhi Shi). Followed by “Handling Ambiguity via Input-Output Kernel Learning“ (authors: Xinxing Xu, Ivor W. Tsang, and Dong Xu), and paper “ConfDTree: Improving Decision Trees Using Confidence Intervals” (authors: Gilad Katz, Asaf Shabtai, Lior Rokach, and Nir Ofek)



Next, the paper titled “Unsupervised Multi-Class Regularized Least-Squares Classification” (authors:  Tapio Pahikkala, Antti Airola, Fabian Gieseke, and Oliver Kramer) was presented. A Novel Semantic Smoothing Method based on Higher Order Paths for Text Classification (authors: Mithat Poyraz, Zeynep Hilal Urhan, and Murat Can Ganiz).


After lunch, I attended the Spatial and Temporal Session. The first paper “Effective and Robust Mining of Temporal Subspace Clusters“ (authors: Hardy Kremer, Stephan Günnemann, Arne Held, and Thomas Seidl). Next, the paper “Student-t based Robust Spatio-Temporal Prediction” (authors: Yang Chen, Feng Chen, Jing Dai, T. Charles Clancy, and Yao-Jan Wu).


Next, Understanding Data Completeness in Network Monitoring Systems (authors: Flip Korn, Ruilin Liu, and Hui (Wendy) Wang). Next, Inferring the Root Cause in Road Traffic Anomalies (authors: Sanjay Chawla, Yu Zheng, and Jiafeng Hu).


During the last session on Tuesday, I attended the “Data Pre-Processing” (see picture below).

On Wednesday morning, Bernhard Schölkopf (Director at Max Planck Institute, see picture below) gave his keynote talk on causal and anticausal learning. He started off by giving the main ideas of causal inference, and discusses implications of underlying causal structures for popular machine learning scenarios.  

Next, I visited a well-attended tutorial on “Mining Uncertain and Probabilistic Data for Big Data Analytics” by Jian Pei (Simon Fraser University, see picture below).

After lunch, it was time for excursions. Three options were available: 1. Magritte museum, 2. Museum of Cocoa and Chocolate, and 3. Cantillon Brewery (Brussels Geuze museum). The second option was the most popular one. Yours truly visited the third option (see pictures below). It’s always nice to have informal talks with conference attendees during such excursions.

Unfortunately, I missed the conference dinner due to technical (server) problems at Ghent University (due to a sudden power outage), which I had to solve. On Thursday morning, the keynote talk was by Jure Leskovec (Stanford University, see picture below) on “Mining Massive Online Networks: Challenges and Opportunities”. He listed several computational approaches how to benefit from the massive amount of social interaction data provided by social networks.

Next, yours truly went to the Classification 3 Session. When I entered the room, the paper “Learning Target Predictive Function without Target Labels” was being presented (authors: Chun-Wei Seah, Ivor Wai-Hung Tsang, Yew-Soon Ong, and Qi Mao, see picture below). The authors introduce a TARget learning Assisted by Source Classifier Adaptation (TARASCA) method. Next, the paper “Fast Kernel Sparse Representation Approaches for Classification” was presented (Yifeng Li and Alioune Ngom).


Next, it was time for the “Learning Attitudes and Attributes from Multi-Aspect Reviews” paper (authors: Julian McAuley, Jure Leskovec, and Dan Jurafsky, see picture below), followed by “Simultaneously Combing Multi-View Multi-Label Learning with Maximum Margin Classification” (authors: Zheng Fang and Zhongfei (Mark) Zhang).


Next, the paper “Decision Theory for Discrimination-aware Classification” (authors: Faisal Kamiran, Asim Karim, and Xiangliang Zhang, see pictures below) was presented.

Next, “Active Label Correction” was presented (authors: Umaa Rebbapragada, Carla E. Brodley, Damien Sulla-Menashe, and Mark Friedl).

...followed by “Sparse Bayesian Adversarial Learning Using Relevance Vector Machine Ensembles” (authors: Yan Zhou, Murat Kantarcioglu, and Bhavani Thuraisingham), and “An AdaBoost Algorithm for Multiclass Semi-Supervised Learning” (authors: Jafar Tanha, Maarten van Someren, and Hamideh Afsarmanesh).


Next, it was time for the 2012 ICDM Community Meeting. First, Prof. Dr. Xindong Wu (Chair of the ICDM Steering Committee, University of Vermont, USA) took the stage. He highlighted that the ICDM community is thriving... highest number of submissions, highest number of attendees, lots of student involvement, ...

... followed by Prof. Dr. Arno Siebes (University of Utrecht, see picture below), and Prof. Dr. Shusaku Tsumoto (Shimane University, Japan, see picture below).

Next, I attended the Feature Selection and Text Mining Session. “Isometric multi-manifold learning for feature extraction” (authors: Mingyu Fan, Hong Qiao, Bo Zhang, and Xiaoqin Zhang) was presented, followed by “Feature Weighting and Selection Using Hypothesis Margin of Boosting” (authors: Malak Alshawabkeh, Javed A. Aslam, Jennifer G. Dy, and David Kaeli).


Next, “An Approach to Evaluate the Local Completeness of Event Logs” (authors: Hedong Yang, Lijie Wen, and Jianmin Wang) was presented, followed by “Cross-Language Opinion Target Extraction in Review Texts” (authors: Xinjie Zhou, Xiaojun Wan, and Jianguo Xiao), “Inductive Model Generation for Text Categorization using a Bipartite Heterogeneous Network” (authors: Rafael Geraldeli Rossi, Thiago de Paulo Faleiros, Alneu de Andrade Lopes, and Solange Oliveira Rezende), “Learning to Refine an Automatically Extracted Knowledge Base using Markov Logic” (authors: Shangpu Jiang, Daniel Lowd, and Dejing Dou), and finally, “Semantic Aspect Discovery For Online Reviews” (authors: Hijbul Alam and SangKeun Lee). See pictures below for the presenters of these papers.


Next, I attended the Panel: Big -- The Value of Data.

Finally, the closing session (see picture below) came with some surprises... the organizers gave away an iPad mini, some Samsung products, and other goodies... Congratulations to the organizing committee for putting together a great event.