# Reinforcement Learning

Reinforcement learning is the process of learning to perform a task by trial and error. Feedback, which can be either positive or negative, is used to inform an agent of the utility of its actions. As the agent repeatedly interacts with its environment, it learns which actions yield to the best outcomes. This type of learning pairs well with many real-world tasks and problems that require a series of decisions.

My work in reinforcement learning was initially focused on learning the fundamentals by implementing and playing with a couple of different problems, all of which use neural networks and the temporal difference algorithm. My thesis was focused on gaining a better understanding of how reinforcement learning behaves under different scenarios by testing in more complex domains. Additionally, I developed abstract and parameterized domains that can take on characteristics of any domain, and I explored the behavior of the temporal difference algorithm using a design of experiments approach to efficiently explore and rigorously analyze characteristics related to the domain, the learning algorithm, and the neural network.

Related publications:

Gatti CJ, 2015.

Design of Experiments for Reinforcement Learning (Springer Theses).*Springer*, ISBN 978-3319121963.

Gatti CJ and Embrechts MJ, 2014.

An application of the temporal difference algorithm to the truck backer-upper problem.*Proceedings of the 22st European Symposium on Artifcial Neural Networks, Computational Intelligence and Machine Learning*, Bruges, Belgium, April 23-25.

Gatti CJ and Embrechts MJ, 2013.

Reinforcement learning with neural networks: Tricks of the trade.*In: Advances in Signal Processing and Machine Learning: Theory and Applications*, vol 410, Springer-Verlag, Berlin-Hiedelberg, pp. 275-310.

Gatti CJ, Embrechts MJ, and Linton JD, 2013.

An empirical analysis of reinforcement learning using design of experiments.*Proceedings of the 21st European Symposium on Artifcial Neural Networks, Computational Intelligence and Machine Learning*, Bruges, Belgium, April 24-26.

Gatti CJ, Linton JD, and Embrechts MJ, 2011.

Reinforcement learning and the effects of parameter settings in the game of Chung Toi.*Proceedings of the 2011 IEEE International Conference on Systems, Man, and Cybernetics (SMC 2011)*, Anchorage, Alaska, October 9-12.

Gatti CJ, Linton JD, and Embrechts MJ, 2011.

A brief tutorial on reinforcement learning: The game of Chung Toi.*Proceedings of the 19th European Symposium on Artifcial Neural Networks, Computational Intelligence and Machine Learning*, Bruges, Belgium, April 27-29.