USNA.edu USNA Computer Science

Home

An Assistant Professor at the United States Naval Academy, I am interested in projects relating to feature selection for value function approximations of MDPs, as well as other topics in reinforcement learning and machine learning.

Any midshipman interested in the fields of Artificial Intelligence, Machine Learning, or Robotics should speak with me; it is likely we could explore some interesting research topics.

Aside from my research, I have a deep interest in teaching and Computer Science education.

Contact Information

Email: taylor at usna.edu
Office: MI-355
Office Phone: 410/293-6816
PGP Fingerprint: BA8F 5FB9 1AE9 BBDD EA28 0F0E 68FF 3C18 DC44 0D5A

Research and Publications

Overview

I am interested in the development and application of value function approximation for Reinforcement Learning. Specifically, I am excited by the promise of methods which automatically select features for linear value function approximation while fitting the value function. One such method, L1-Regularized Approximate Linear Programming (RALP) is of interest as it is the only method to perform automated feature selection while approximating the optimal value function from off-policy samples.

Interesting applications I am exploring include the use of value function information in human performance analysis, human-supervised autonomy, and anomaly detection in autonomous systems.

Curriculum Vitae

Refereed Conference Papers

* Midshipman co-author

Student Research

Dissertation

Workshop Papers

Teaching

Fall AcYear 2015
SY301: Data Structures for Cybersecurity [ web ]

Spring AcYear 2014
SI475: Intelligent Robotics [ web ]
SI496C: Machine Learning

Fall AcYear 2014
IC312: Data Structures [ web ]

Spring AcYear 2013
SI475: Intelligent Robotics
IC211: Object-Oriented Programming
SI486A: Machine Learning

Fall AcYear 2013
IC312: Data Structures

Spring AcYear 2012
IC211: Object-Oriented Programming

Fall AcYear 2012
IC210: Introduction to Computer Programming

Before USNA
CS196s: Teaching with Robots (Duke University, Fall 2010)

Calendar

This calendar is provided to ease the scheduling of EI and advising appointments.

The best way to make an appointment with me is to find a time not marked busy on this calendar which works for you, and e-mail me.

Thursdays are my "research day," and I try to keep them relatively distraction free. However, if you need EI, you are welcome to ask for a time on Thursday.