This package contains methods and algorithms for applying the method of moments to learn stochastic languages (e.g., WA, HMMs, PSRs). The package provides reference implementations for the current (as for 2014) state-of-the-art in moment-based learning for stochastic languages and also contains the benchmark PAutomaC dataset. The code (and link to the accompanying paper) is (available on GitHub).
This package is a collection of Java tools and methods for using predictive state representations (PSRs) in order to make predictions and construct model-based reinforcement learning agents. The code uses many optimizations (e.g., randomized compression) in order to improve efficiency and is (available on GitHub).
If you have an questions about these packages please send me an email, and I will do my best to clear things up. However, please be aware that I am no longer working on these packages directly (though I will try to be responsive with requested fixes etc.). And, as always, you are warned that these are open-source packages intended for research use; neither I, nor my co-authors, can take any responsibility if you somehow misuse them and wreak havoc on yourself or others.