We spoke on Skype this week. We discussed how to edit Weka in Eclipse, including modifications to the classpath. We went over the explanation of M-estimate smoothing, as opposed to Laplace smoothing, and Dr. Dutta asked me to implement the M-estimate in Weka.
We discussed different ranking schemes. Dr. Dutta asked me to read more about the ROC curve and AUC, as well as the Brier score. We went over the paper "Learning to Rank: From Pairwise to Listwise Approaches," putting it in context with what we are doing.
Dr. Dutta assigned further readings and asked me to update the minutes and related literature on this site.
I found the Laplace module in the Weka source files, and created duplicate functions implementing the M-estimate. This was my first experience adding my own code to a project of this size, and I enjoyed exploring the Weka code and the challenge of fitting my own in.
Reading about the different scoring methods was very interesting, as was Dr. Dutta's explanation of the paper on listwise ranking. The project is coming together, and I am beginning to feel that I have a working understanding of this area of machine learning.