Amber's Research
Most of the links below relate to my research on Pattern Recognition, but I am also working on projects related to Respondent Driven Sampling and Network Modelling. I hope to have more information about these projects available shortly.My current work on Pattern Recognition is focussed on the design of Multiple Classifier Systems (also known as ensemble methods, among other names), with a particular focus on methods for non-stationary classification. I am currently trying to put together an R package of methods for non-stationary classification using MCSs - if you have an algorithm you would like included, or would like to help, please let me know! I am also interested in developing methods and tools to help with the comparison and evaluation of different algorithms.
Peer-reviewed Publications
A. Tomas (2011) A Dynamic Logistic Multiple Classifier System for Online Classification Proc. 10th International Workshop on Multiple Classifier Systems Naples, 2011, LNCS 6713, pdf The original publication is available at www.springerlink.com
A. Tomas & Krista J. Gile (2011) The Effect of Differential Recruitment, Non-response and Non-recruitment on Estimators for Respondent Driven Sampling. Electronic Journal of Statistics 5 899-934. link
A. Tomas (2009) Constraints in Weighted Averaging, Proc. 8th International Workshop on Multiple Classifier Systems, Reykjavik, 2009, LNCS 5519, 354-363, pdf The original publication is available at www.springerlink.com
A. Tomas (2008) Combining Methods for Dynamic Multiple Classifier Systems, Proc. 3rd International Workshop on Artificial Neural Networks in Pattern Recognition, Paris, 2008, LNAI 5064, 180-192, pdf The original publication is available at www.springerlink.com
Other Publications
A. Tomas (2008) A Dynamic Logistic Model for Combining Classifier Outputs, DPhil Thesis, The University of Oxford (supervisor Prof B.D. Ripley) abstract, fulltext
A. Tomas (2003) Aspects of Bayesian Networks, Honours thesis, The University of Adelaide pdf
Conference and Seminar Presentations
The Design of Multiple Classifier Systems, Department of Statistics Seminar Series, The University of Oxford (2009) pdf
Constraints in Weighted Averaging, The 8th International Workshop on Multiple Classifier Systems, Reykjavik (2009) pdf
Extending Multiple Classifier Systems to Dynamic Scenarios, Research Student Seminar, Department of Statistics, The University of Oxford (2008) pdf
Implementation of a Dynamic Logistic Model for Classification, Annual conference of the German Classification Society (GfKl), Hamburg (2008) pdf
Dynamic Mixture Models for Classification, Annual conference of the German Classification Society, Freiburg (2007) pdf
Classification: When is Temple Cowley not Temple Cowley? Lady Margaret Hall Mathematics Seminar Series (2006) pdf
Decision Making and Learning using Bayesian Networks, Oxford Graphical Models Group (2005) pdf
Posters
Combining Methods for Dynamic Multiple Classifier Systems. Presented at The Third International Workshop on Artificial Neural Networks in Pattern Recognition (ANNPR), Paris 2008. pdfBits and Bobs
I strongly believe that sharing ideas and information is the best way for research to progress. And progress is a good thing! With that in mind, here is my small contribution - some thoughts and notes on various topics which I haven't published elsewhere but may be of interest to someone out there. If you find any of this material particularly useful, or would like more information, please let me know!
Notes on the Relationships between Common Combining Rules for Multiple Classifier Systems pdf
Respondent Driven Sampling has previously been modelled as a random walk on a network. This document shows that the model can be used to encompass within-group differential recruitment, and examine the implications for bias of several common estimators. pdf
More coming...