I am a researcher in the Machine Learning and Applied Statistics group at Microsoft Research. For a while, I was looking at applications of machine learning in speech recognition and related problems in the Speech group.

Before joining Microsoft in 2001, I completed a Ph.D. at the Center for Language and Speech Processing at Johns Hopkins University in Baltimore. Before that, I recieved my B.S. in Electrical Engineering from Lafayette College in Easton, PA.

I was born in Kandy, Sri Lanka, where I grew up, except for stints in Kyoto, Maine, and Virginia. In my free time, I run, cook, ski, hike, rock climb, and enjoy the Pacific Northwest. Check out my blog.

PhD thesis

Gunawardana, A. (2001).  The Information Geometry of EM Variants for Speech and Image Processing.  Ph.D. Dissertation.  The Johns Hopkins University, Baltimore, MD. (pdf|ps)

Journal papers

Gunawardana, A. and Byrne, W. (2005).  Convergence theorems for generalized alternating minimization procedures.   Journal of Machine Learning Research.  6(Dec), 2049--2073. 

Gunawardana, A. and Byrne, W. (2001).  Discounted likelihood linear regression for rapid speaker adaptation.   Computer, Speech and Language.  15(1), 15--38. 

Byrne, W. and Gunawardana, A. (2000).  Comments on "Efficient training algorithms for HMM's using incremental estimation."  IEEE Transactions on Speech and Audio Processing.  8(6), 751--754.  Correspondence. 

Dougherty, A. and Gunawardana, A. (1994).   "The Mean Shape of Three-Dimensional Dendrites: A Comparison of Pivalic Acid and Ammonium Chloride,"    Physical Review E.   50, 1349--1352.

Refereed conference and workshop papers

Gunawardana, A. and Meek, C. (2008).  Aggregators and Contextual Effects in Search Ad Markets.   In WWW Workshop on Targeting and Ranking for Online Advertising. (slides)

Mahajan, M., Gunawardana, A., and Acero, A. (2006).  Training algorithms for hidden conditional random fields.   In International Conference on Acoustics, Speech, and Signal Processing.

Gunawardana, A., Mahajan, M., Acero, A, and Platt J. C. (2005).  Hidden conditional random fields for phone classification .  In International Conference on Speech Communication and Technology.

Li, X., Gunawardana, A. and Acero, A. (2005).  Unsupervised semantic intent discovery from call log acoustics.   In International Conference on Acoustics, Speech, and Signal Processing, 45--48.  

Gunawardana, A. and Acero, A. (2003).  Adapting acoustic models to new domains and conditions using untranscribed data.  In International Conference on Speech Communication and Technology.  

Gunawardana, A. and Byrne, W. (2001).  Discriminative speaker adaptation with conditional maximum likelihood linear regression.  In International Conference on Speech Communication and Technology, 1203--1206. 

Gunawardana, A. and Byrne, W. (2001).  Convergence of DLLR rapid speaker adaptation algorithms.  In ISCA-ITR Workshop on Adaptation Methods for Speech Recognition, 45--48. 

Gunawardana, A. and Byrne, W. (2000).  Robust estimation for rapid speaker adaptation using discounted likelihood techniques.  In International Conference on Acoustics, Speech, and Signal Processing, 985--988. 

Byrne, W. and Gunawardana, A. (1999).  Discounted likelihood linear regression for rapid adaptation.  In European Conference on Speech Communication and Technology, 203--206. 

Byrne, W. and Gunawardana, A. (1999).  Convergence of EM variants.  In IEEE Information Theory Workshop on Decisions, Estimation, Classification, and Imaging.

Technical Reports

Gunawardana, A. (2001).  "Maximum Mutual Information Estimation of Acoustic HMM Emission Densities," CLSP Research Note No. 40. Center for Language and Speech Processing, Johns Hopkins University, Baltimore, MD.