website/content/research/deepreinforcementlearning/_index.md

25 lines
1.4 KiB
Markdown
Raw Normal View History

2020-01-15 21:51:49 -05:00
---
Title: Deep Reinforcement Learning
Description: Combining Reinforcement Learning with Deep Learning
---
2022-02-17 13:37:46 -05:00
I am interested in sample-efficient reinforcement learning.
That is, decreases the number of interactions an agent needs
with an environment to achieve some goal. In the Fall of 2019,
I approached this by integrating interactive demonstration
data into the optimized Deep Q-Networks algorithm.
2020-01-15 21:51:49 -05:00
The results are positive and are heavily documented through the following:
2022-02-17 13:37:46 -05:00
[Undergraduate Honors Thesis](/files/research/honorsthesis.pdf)
2020-01-15 21:51:49 -05:00
2022-02-17 13:37:46 -05:00
[Undergraduate Honors Defense](/files/research/ExpeditedLearningInteractiveDemo.pptx)
2020-01-15 21:51:49 -05:00
Thanks to my advisor Dr. Ron Zacharksi and my committee members for all their feedback on my work!
2022-02-17 13:37:46 -05:00
The semester prior, I built a [reinforcement learning library](https://github.com/brandon-rozek/rltorch) with implementations of several popular papers. ([Semi-Weekly Progress](weeklyprogress)).
2020-01-15 21:51:49 -05:00
2022-02-17 13:37:46 -05:00
I also presented at my school's research symposium. ([Slides](/files/research/QEP.pptx)) ([Abstract](abstractspring2019))
2020-01-15 21:51:49 -05:00
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](https://github.com/brandon-rozek/zerogym) and [HTTP](https://github.com/brandon-rozek/gymhttp). Given the option, you should use the ZeroMQ implementation since it contains less communication overhead.