Visiting Assistant Professor of Computer Science
At Berea College since 2021
Contact Information
Danforth Technology Building, 102B
CPO 2188
Email: shepherdp@berea.edu
Spring 2022
Office Hours
Mon: 11:00 a.m. – noon
In-person
Tue: noon – 2:00 p.m.
In-person
Fri: 11:00 a.m. – noon
In-person
Class Schedule
- CSC 226 A (Mon/Wed/Fri: 1:20 pm – 2:30 pm)
- CSC 226 B (Mon/Wed/Fri: 9:20 am – 10:30 am)
- CSC 486 PS (Mon/Wed/Fri: 12:00 pm – 1:10 pm)
Education
- University of Kentucky, Ph.D. Computer Science, current GPA: 4.0, 2015-2021
- University of Kentucky, B.S. Business and Economics, Minor: Computer Science, Magna Cum Laude, 2013-2015
- Bluegrass Community and Technical College, Associate in Science, High academic distinction, 2011-2013
Interviews
TechGuide.org interview about the pros and cons of a Master’s in AI: https://techguide.org/computer-science/masters-in-artificial-intelligence/#expert=patrick-shepherd
Publications
Shepherd, P., Weaver, M., and Goldsmith, J. (2020). ‘An Investigation into the Sensitivity of Social Opinion Networks to Heterogeneous Goals and Preferences.’ International Symposium on Foundations and Applications of Big Data Analytics. The Hague, Netherlands.
Shepherd, P. and Goldsmith, J. (2020). ‘A reinforcement learning approach to strategic belief revelation with social influence.’ Thirty-fourth AAAI Conference on Artificial Intelligence. New York.
Shepherd, P., Weaver, M., and Goldsmith, J. (2020). ‘Opinion Diffusion Software with Strategic Opinion Revelation and Unfriending.’ arXiv preprint arXiv:2006.12572.
Manuscripts in Progress
Shepherd, P., Higashi, R., Fan, T. W.-M., and Lane, A. (under revision) ‘IsoMap: A Graph-based Algorithm for Assignment of Mass Spectra from Multiply-Enriched Biological Samples.’ Unpublished.
Software in Use
PREMISE: Precalculated Exact Mass Isotopologue Search Engine (currently closed-source). A software suite for curation and compilation of molecular profiles in mass spectra from singly- or multiply-enriched biological samples.
Service
- Reviewed for Energy Reports
- Reviewed for AISTATS
- Reviewed for AIES
- Reviewed for PLOS One
Current Research
My current research deals with using multi-agent reinforcement learning (MARL) in social network frameworks to develop policies for interaction. I have built a flexible and powerful social network model that is able to replicate the research conditions of a great deal of peer-reviewed research in opinion diffusion, epidemiology, and voting theory. Where these fields tend to view networks as static, my framework highlights the dynamism of social networks. This framework is a custom-built environment for learning agents, in which each agent has access to a menu of actions they can use to interact with others in the network to whom they are connected. I am currently developing novel algorithms to allow agents to learn the best way to behave in this environment given their available actions in order to maximize a personal sense of utility.
The body of my current research focuses on two broad questions: how to induce the most effective learning of this ilk in the agents and what sorts of behavior they tend towards, and what the effects of this learned behavior are on the network as a whole.
Research Interests
- Artificial intelligence
- Machine learning
- Reinforcement learning
- Dynamics of social network information flow
- Structural social network analysis
Honors and Memberships
- Burt L. Sims Graduate Fellowship – September 2017
- Thaddeus B. Curtz Memorial Scholarship – April 2016
- Omicron Delta Kappa – National Honor Society
- Phi Theta Kappa – National Honor Society
- Dean’s List: August 2012 – May 2015