I work as data scientist with a start-up company, I have a linear algebra/measure theory heavy math background (I did some graduate level course work in those subjects while doing my bachelor's degree). I've spent a decent amount of time studying Neural Networks when I got this job (I also do a lot of non-ML algorithmic work for the company) but I've realized that the best results in the field I'm working in use techniques like Hidden Markov Support Vector Machines.
I've been using "Predicting Structured Data" to get a better handle on the mathematics of using Kernel methods mixed with Probabilistic Graphical Models. I've found it does a great job of covering the material, but it doesn't have much in the way of exercises - I'm just left trying to prove propositions on my own, etc.
Is there a better or more popular book that covers this material?
Link: http://mitpress.mit.edu/books/predicting-structured-data
[link][1 comment]