My research interests include machine learning. My advisor is Thorsten Joachims. My area of research is learning functions with structured outputs in domains requiring approximation inference algorithms. When using approximations many of the theoretical gaurantees of structured learning algorithms no longer hold and in practice the learned models are of inconsistent quality, and my work seeks to provide new theory and empirical understanding to this area.
My graduate work has included supervised clustering, loopy MRF learning, and data mining of the survey data from Project FeederWatch.
While I was an undergrad at Duke, I worked with Dr. Susan Rodger on computer science education visualization tools. This included in 2001 the JAWAA editor, and in 2002 onward the totally new JFLAP 4.0!
Here are my published documents in reverse chronological order:
T. Finley, T. Joachims, Supervised k-Means Clustering, Under submission, 2008.
[PDF]
T. Finley, T. Joachims, Training Structural SVMs when Exact Inference is Intractable, ICML (to appear), 2008.
[PDF]
T. Joachims, T. Finley, C.-N. J. Yu, Cutting-plane training of structural SVMs., Machine Learning (to appear), 2007.
[PDF]
T. Finley, T. Joachims, Parameter Learning for Loopy Markov Random Fields with Structural Support Vector Machines, ICML Workshop on Constrained Optimization and Structured Output Spaces, 2007.
[PDF]
Yisong Yue, T. Finley, F. Radlinski, T. Joachims, A Support Vector Method for Optimizing Average Precision, Proceedings of the Conference on Research and Development in Information Retrieval (SIGIR), 2007.
[PDF]
Susan Rodger and Thomas Finley, JFLAP - An Interactive Formal Languages and Automata Package, ISBN 0763738344, Jones and Bartlett, published 2/27/06, 2006.
[Info]
Susan H. Rodger, Bart Bressler, Thomas Finley, and Stephen Reading, Turning Automata Theory into a Hands-on Course, Thirty-seventh SIGCSE Technical Symposium on Computer Science Education, 2006.
T. Finley and T. Joachims, Supervised Clustering with Support Vector Machines, Proceedings of the International Conference on Machine Learning (ICML), 2005.
[Postscript] [PDF]
Ryan Cavalcante, Thomas Finley and Susan H. Rodger, A Visual and Interactive Automata Theory Course with JFLAP 4.0, Thirty-fifth SIGCSE Technical Symposium on Computer Science Education, 2004 (p.140-144).
[PDF]
Ayonike Akingbade, Thomas Finley, Diana Jackson, Pretesh Patel and Susan H. Rodger, JAWAA: Easy Web-Based Animation from CS 0 to Advanced CS Courses, Thirty-fourth SIGCSE Technical Symposium on Computer Science Education, p. 162-166, 2003.
[PDF]
| Thomas Finley | • | Home | About Me | Research | Projects | Notes |