School of Electrical Engineering and Computer Science

Oregon State University

Explorative data clustering involves grouping objects into clusters such that similar objects are grouped together. My research attemps to advance the field of unsupervised clutsering in a number of directions.
First, motivated by the fact that objects in a data set maybe similar to each other in multiple different ways, and different clustering structures may exist in the same data. I am interested in exploratively examining data in different ways to produce different clusterings. Such clusterings can be sometimes combined to provide a more reliable view of the structure of the data via cluster ensemble methods, or other times examined individually as they may provide different insights (non-redundant clustering).

Clustering with partial supervision and active learning.  Due to the inherent embiguous nature of the clustering task, in practice it is very useful to consider some user-provided side information such that a learning algorithm can seek to find a underlying clustering struture to that is most consistent with the side information. For example such information can be expressed in terms of pairwise constraints requiring some objects to be placed together or apart. I am interested in developing clustering techniques that can take into consideration richer forms of user constraints (e.g., comparative constraints involving multiple objects), and active learning frameworks that effectively acquire various form of user inputs without imposing heavy burden on the users. This work is supported by  my NSF career grant. For more details, please see the project page here.


Selected publications:

Wei Zhang, Akshat Surve, Xiaoli Z. Fern and Thomas Ditteriech, Learning Non-redundant Codebooks for Classifying Complex Objects,  In Proceedings of International Conference on Machine Learning, ICML 2009. PDF

Javad Azimi and Xiaoli Fern, Adaptive Cluster Ensemble Selection, In Proceedings of International Joint Conference on Artificial Intellegence, IJCAI 2009. PDF

Xiaoli Z. Fern and Wei Lin, Cluster Ensemble Selection, Journal of Statistical Analysis and Data Mining, Special Issue on Best of SDM08, Volume 1, Issue 3 , Pages128 - 141, 2008 Preprint

Xiaoli Z. Fern and Wei Lin, Cluster Ensemble Selection, In Proceedings of 2008 SIAM International Conference on Data Mining (SDM08). pdf

Ying Cui, Xiaoli Z. Fern and Jennifer Dy, Non-redundant multi-view clustering via orthogonalization, in Proceedings of 7th IEEE International Conference on Data Mining (ICDM07) pdf.

Xiaoli Z. Fern and Carla E. Brodley,  "Cluster ensembles for high dimensional data clustering: An empirical study", Techenical report  CS06-30-02.

Xiaoli Z. Fern and Carla E. Brodley,  "Solving cluster ensemble problems by bipartite graph partitioning", in Proceedings of 21th International Conference on Machine learning (ICML2004), PDF file, Matlab implementation of the algorithm ( Note: this code is provided on "as is" basis for research use only. )

Xiaoli Z. Fern and Carla E. Brodley,  "Random Projection for High Dimensional Data Clustering: A Cluster Ensemble Approach",  in Proceedings of 20th International Conference on Machine learning (ICML2003), PDF file

• The Bird-Bioacoustic Project. In this project, we record and analyze bird songs to predict bird species and to understand the behavior of birds through their songs. This is a joint project with Dr. Raviv Raich (ECE, OSU) and Dr. Matthew Betts (Forrestry, OSU). See project webpage for more exciting details and publications related to this topic.
  


• Forrest Briggs, Xiaoli Fern and Raviv Raich, Context-Aware MIML Instance Annotation, To appear in Proceedings of IEEE International Conference on Data Mining (ICDM 2013)

• Forrest Briggs, Xiaoli Fern and Raviv Raich, Rank Loss Support Instance Machines for MIML Instance Annotation, In Proceedings of ACM International Conference on Knowledge Discovery and Data Mining (KDD 2012)

• Forrest Briggs, Balaji Lakshminarayanan, Lawrence Neal, Xiaoli Z. Fern, Raviv Raich, Sarah Frey, Adam Hadley, and Mattew G. Betts, Acoustic classification of multiple simultaneous bird species: A multi-instance multi-label approach, To appear in Journal of Acoustics Society of America, 2012

• Lawrence Neal, Forrest Briggs, Raviv Raich and Xiaoli Fern, Time-frequency segmentation of bird song in noisy acoustic environments, to appear in Proceedings of the 36th International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2011

• Forrest Briggs, Raviv Raich, and Xiaoli Z. Fern, Audio Classification of Bird Species: a Statistical Manifold Approach, to appear in Proceedings of International Conference on Data Mining, ICDM 2009, pdf

• B. Lakshminarayanan, R. Raich, and X. Fern, A syllable-level probabilistic framework for bird species identification, in Proc. IEEE International Conference on Machine Learning and Applications, 2009.

• Experimental design and Bayesian Optimization . In this project, we develop machine learning techniques to help scientists and engineers to desgin better microbial fuels by allowing them more efficiently experiment with different nano-structures. This is an experimental design problem --- how to choose appropriate experiments to perform under a budget constraint.

• Ali Jalali, Javad Azimi, and Xiaoli Fern, A Lipschitz Exploration-Exploitation Scheme for Bayesian Optimization, In Proceedings of ECML/PKDD 2013

• Javad Azimi, Ali Jalali, and Xiaoli Fern, Hybrid Batch Bayesian Optimization, In Proceedings of International Conference on Machine Learning (ICML 2012)

• Javad Azimi, Alan Fern and Xiaoli Fern, Batch Active Learning via Coordinated Matching, In Proceedings of International Conference on Machine Learning (ICML 2012)
  
  
• Javad Azimi, Alan Fern and Xiaoli Fern, Budgeted Optimization with Concurrent Stochastic-Duration Experiments, In NIPS 2011 (spotlight)

• Javad Azimi, Alan Fern and Xiaoli Fern, Batch Bayesian Optimization via Simulation Matching Advances in Neural Information Processing Systems (NIPS-2010)PDF

• Javad Azimi, Xiaoli Fern, Alan Fern, Elizabeth Burrows, Frank Chaplen, Yanzhen Fan, Hong Liu, Jun Jiao and Rebecca Schaller. Myopic Policies for Budgeted Optimization with Constrained Experiments In Proceedings of AAAI Conference on Artificial Intelligence (AAAI-2010) PDF.

• Mining human strategies (behavioral patterns) from human computer interaction data . In this research, we apply data mining to HCI log data with the goal of finding interesting behavioral patterns that shed some lights on the strategies users employ while using software for problem solving. We collaborate with the HCI researchers at EUSES consortium on this project.


• Xiaoli Fern, Chaitanya Komireddy, Valentina Grigoreanu, and Margaret Burnett, "Mining Problem-Solving Strategies from HCI Data", In ACM Transaction on CHI (TOCHI) 2010.
  

• Neeraja Subrahmaniyan, Laura Beckwith, Valentina Grigoreanu, Margaret Burnett, Susan Wiedenbeck,Vaishnavi Narayanan, Karin Bucht, Russell Drummond, and Xiaoli Fern, “Testing vs. Code Inspection vs. ... What Else? Male and Female End Users’ Debugging Strategies”, ACM Conference on Human-Computer Interaction, April 2008  (CHI2008)
  

• Xiaoli Fern, Chaitanya Komireddy, Margaret Burnett, "Mining Interpretable Human Strategies: A Case Study", to appear in Proceedings of 7th IEEE International Conference on Data Mining (ICDM07) pdf. A longer tech report version can be found here pdf.

• Grigoreanu, V., Beckwith, L., Fern, X., Yang, S., Komireddy, C., Narayanan, V., Cook, C., Burnett, M., Gender Differences in End-User Debugging, Revisited: What the Miners Found. In Proceedings of IEEE Symposium on Visual Languages and Human-Centric Computing Languages and Environments, Brighton, England, September 2006. pdf

• Correlation analysis on Earth science data. In this project, the goal is to find out how one environmental factor impact another, for example Sea Surface Temperature (SST) versus Sea Surface Pressure (SSP), and vegetation index versus precipitation. We developed a method that builds mixture of linear canonical correlation (CCA) models to explain correlation patterns between different domains.
• Xiaoli Z. Fern, Carla E. Brodley and Mark A. Friedl,  "Correlation clustering for learning mixture of canonical correlation models ",  SIAM International Conference on Data Mining (SDM2005). PDF file; Matlab code for generating synthetic data with a prespecified correlation structure.