Deep Reinforcement Learning

Brandon Rozek

In the Fall of 2019, I look at integrating demonstration data into a reinforcement learning algorithm in order to make it sample efficient.

The results are positive and are heavily documented through the following:

Honors Thesis

Honors Defense

Thanks to my advisor Dr. Ron Zacharksi and my committee members for all their feedback on my work!

In the spring of 2019, under the guidance of Dr. Ron Zacharski I practiced several of the modern techniques used in Reinforcement Learning today.

I facilitated my learning by creating a reinforcement learning library with implementations of several popular papers. (Semi-Weekly Progress)

I also presented my research (which involved creating an algorithm) at my school’s research symposium. (Slides) (Abstract)

In the summer of 2019, I became interested in having the interactions with the environment be in a separate process. This inspired two different implementations, ZeroMQ and HTTP. Given the option, you should use the ZeroMQ implementation since it contains less communication overhead.