Generalization in RL
Project Description
Overfitting occurs in Reinforcement Learning when agents memorize action sequences instead of learning an actual skill or mistakenly correlate rewards with certain spurious features from the observations generated by the Markov Decision Process (MDP).
Keywords
Reinforcement Learning
, Generalization
, Decoupled Networks
Methods
a) Attention-based Partial Decoupling of Policy and Value (APDAC)
In this project, we build on the works of IDAAC but partially separate the policy and the value function for better generalization
b) Hyperbolic Discounting in Generalization
In this project, we explore the benefits of hyperbolic discounting in generalization.
Current Team Members
- Nasik Muhammad Nafi - Ph.D. student, Computer Science, Kansas State University - (Lead)
- Raja Farrukh Ali - Ph.D. student, Computer Science, Kansas State University -
- Kevin Duong - Undergraduate, Computer Science, Kansas State University
Affiliates
- None
Alumni
- Creighton Glasscock - BS Computer Science 2022 {now at Michigan State University pursuing MS}
Data Sets
TBD
Link to any data sets for the project here. For any federally-sponsored research project (especially NSF and NIH-sponsored projects) of the KDD Lab, there must be an open access data repository. These may be subdirectories of a Bitbucket repository, but link to them separately here anyway. For all other projects, link to sites where data produced from the project are shared - e.g., a publisher server, Kaggle landing page, NIST documentation page, etc. Link to any data sharing agreements here.
Trello Board
TBD Every KDD Lab project must have a Trello Team and Trello Board, which must be private. Link to the Trello Board for the project here.
Source Code
TBD Every KDD Lab project must have a Bitbucket repository, which may be public or private. Link to the repository or repositories for the project here.
References
TBD
Background and Related Work
TBD
See the Machine Learning and Probabilistic Reasoning subpages of the original KDD wiki (v1, 2001 - 2006).
KDD Lab Publications
TBD
Use APA citation format and make sure citations are synchronized with the pages listing conference papers, journal articles, book chapters, posters, and student publications.
- De La Torre, M. F., Aguirre, C. A., Anshutz, B., & Hsu, W. (2018). MATESC: Metadata-Analytic Text Extractor and Section Classifier for Scientific Publications. Proceedings of the 10th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2018): International Conference on Knowledge Discovery and Information Retrieval (KDIR 2018), Seville, Spain, September 18-20, 2018
- Yates, H., Chamberlain, B., Healey, J., & Hsu, W. (2018). Binary Classification of Arousal in Built Environments using Machine Learning. Working Notes of the 2nd International Joint Conference on Artificial Intelligence (IJCAI) Workshop on Artificial Intelligence in Affective Computing, Stockholm, Sweden, July 15, 2018.
Last updated Thursday, Nov 26th, 2022
Last updated by pozegov on Jul 6, 2023