Jan-Feb Wrap Up

Basically, I have been focusing on Probability and Statistics since I need a better grasp on this area of mathematics and its application to Machine Learning, in particular Bayesian Inference.  I am familiar with Bayes Theorem, but I am not familiar with Bayesian Belief and Markov Belief Networks, a topic covered in Judea Pearl’s book Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference.

I reviewing some other video lectures in Stanford University’s YouTube such as CS106A Program Methodology taught by Professor Mehran Sahami.  In addition I have reviewing a couple of MIT OpenCourse courses such as Introduction to Probability and Statistics, and Mathematics for Computer Science.

Reinforcement Learning

Reinforcment learning (RL) is a branch of machine learning (ML) in which an agent takes action based on maximizing a reward function.  The problem area includes such as the n-arm bandit problem, balancing pole problem and temporal difference learning.  Other names for reinforcment learning are sequential decision making and adaptive control.  The mathematics is based on Markov Decision Process (MDP) and adaptive control theory.   Recent work in reinforcement learning includes research in Bayesian approaches along with Mutli-agent reinforcement learning (MARL).

To begin your RL exploration, start with the RLAI web site, and follow the hyperlink to RL book by Rich Sutton.