Zhuang (John) Wang

 

IBM

Global Business Services – Business Analytics and Optimization
Email: firstname@temple.edu

Research Interests

Large-scale data classification

Support Vector Machine

Online learning

Multiple Instance Learning

 

My current work is focused on bridging science and business by designing and developing innovative big data analytics solutions that leverage heterogeneous data from multiple sources for business innovations. 

Education

Ph.D., Computer and Information Sciences (Machine learning and Data mining), Temple University, Philadelphia, USA, 2006~2010.

B.A., Electronic Commerce, Wuhan University, Wuhan, China, 2002~2006.   

Selected Publications

1.        BudgetedSVM: A Toolbox for Large-scale Non-linear SVM [open source software]

Paper under review. This toolbox is designed for training non-linear SVM on large-scale, high-dimensional data when it cannot
fit into memory. It can be treated as a missing link between LibLinear and LibSVM, combining efficiency of linear SVM with accuracy of kernel SVM models.

 

1.        Log-based Predictive Maintenance

Sipos, R., Fradkin, D., Moerchen, F. and Wang, Z.

Machine Learning Journal (MLJ), accepted (minor revision).

 

2.        Large-scale Non-linear Classification: Algorithms and Evaluations

Wang, Z.

Int. Joint Conf. on Artificial Intelligence (IJCAI), to be in Tutorial. [tutorial homepage]

 

3.        Breaking the Curse of Kernelization: Budgeted Stochastic Gradient Descent for Large-Scale SVM Training [pdf] [codes]

Wang, Z., Crammer, K and Vucetic, S.

Journal of Machine Learning Research (JMLR), 2012.

 

4.        Scaling up Kernel SVM on Limited Resources: a Low-rank Linearization Approach [pdf]

Zhang, K., Lan, L., Wang, Z., and Moerchen, F.

Int. Conf. on Artificial Intelligence and Statistics (AISTATS), 2012.

 

5.        Mixture Model for Multiple Instance Regression and Applications in Remote Sensing [pdf] [data]

Wang, Z., Lan, L. and Vucetic, S.

IEEE Transactions on Geoscience and Remote Sensing (TGRS), 2012.

 

6.        Trading Representability for Scalability: Adaptive Multi-Hyperplane Machine for Nonlinear Classification [pdf] [codes]

Wang, Z., Djuric, N., Crammer, K. and Vucetic, S.

 ACM SIGKDD Int. Conf. on Knowledge Discovery and Data Mining (KDD), 2011, (Oral presentation AR: 56/714 = 7.8%)

 

7.        An Active Learning Algorithm Based on Parzen Window Classification [pdf]

 Lan, L., Shi, H., Wang, Z. and Vucetic, S., 

Journal of Machine Learning Research (JMLR): W&C Proc. 2010 AISTATS Active Learning Challenge, (our results ranked as the 5th in the competition)

 

8.        Multi-Class Pegasos on a Budget [pdf]

Wang, Z., Crammer, K and Vucetic, S.  

Int. Conf. on Machine Learning (ICML), 2010

 

9.        Online Training on a Budget of Support Vector Machines using Twin Prototypes [pdf] 

Wang, Z. and Vucetic, S.

Statistical Analysis and Data Mining Journal (SAM), 2010.

 

10.    Online Passive-Aggressive Algorithms on a Budget [pdf]   

Wang, Z. and Vucetic, S.

Int. Conf. on Artificial Intelligence and Statistics (AISTATS), 2010.

 

11.    Fast Online Training of Ramp Loss Support Vector Machines [pdf] 

Wang, Z. and Vucetic, S.,

IEEE Int. Conf. on Data Mining (ICDM), 2009, (Oral presentation AR: 70/786 = 8.9%)

 

12.    Compressed Kernel Perceptrons [pdf] 

Vucetic,S., Coric, V. and Wang, Z.,

IEEE Data Compression Conf. (DCC), 2009, (Oral presentation)

 

13.    Twin Vector Machines for Online Learning on a Budget [pdf]

Wang, Z. and Vucetic, S. 

SIAM Conf. on Data Mining (SDM), 2009, (Oral presentation AR: 55/351 = 15.67%)

 

14.    Aerosol Optical Depth Prediction from Satellite Observations by Multiple Instance Regression [pdf]

Wang, Z., Radosavljevic, V., Han, B., Obradovic, Z. and Vucetic, S.

SIAM Conf. on Data Mining (SDM), 2008, (Oral presentation AR: 40/282 = 14.18%)

         

 

Last Update: Apr-26-2013