Maximum Likelihood Learning and Inference in Conditional Random Fields is the thesis I wrote during my undergraduate degree at University of Copenhagen under the supervision of Christian Igel. The thesis centres around the class of models called Conditional Random Fields, – a new and now very popular model in machine learning research – which has proven to be very effective for image processing (object recognition, image segmentation and image denoising) and natural language data processing (e.g. POS tagging). The thesis develops a mathematical framework for the models, different procedures for training them and implements a very fast method for performing exact inference.
Maximum Likelihood Learning and Inference in Conditional Random Fields, Thesis (PDF)
The corrected thesis produced by Iulian Vlad Serban at the Department of Mathematical Sciences, University of Copenhagen.
Software Program (ZIP)
The corrected software program developed and used in the thesis. It has been developed for Ubuntu 12.04, but should compile for other platforms as well. The software is subject to the disclaimer found in the zip file