Jeff Johns, Postdoctoral Research Fellow, Department of Computer Science, Duke University

Duke University
Department of Computer Science
Box 90129
Durham, NC 27708

Office: D224 LSRC
Lab Phone: (919) 660-6553

johns [at] cs.duke.edu

News

Recipient of a Computing Innovation Fellowship sponsored by the Computing Community Consortium and the Computing Research Association with funding from the National Science Foundation. My mentor for this two-year fellowship is Professor Ronald Parr of Duke University's Department of Computer Science.

Graduated with a Ph.D. and M.S. in Computer Science from the University of Massachusetts Amherst. I was a member of the Autonomous Learning Laboratory and my advisor was Professor Sridhar Mahadevan.

Postdoctoral Research Fellow
Computer Science, Duke University, 2010-2011

Ph.D. and M.S. in Computer Science
University of Massachusetts Amherst, 2010

B.S. with Distinction in Chemical Engineering
University of Virginia, 1998

Jeff Johns.
Basis Construction and Utilization for Markov Decision Processes using Graphs.
PhD thesis, Computer Science, University of Massachusetts Amherst, 2010.
[.pdf] [.pptx]

Jeff Johns, Christopher Painter-Wakefield, & Ronald Parr.
Linear Complementarity for Regularized Policy Evaluation and Improvement.
Advances in Neural Information Processing Systems (NIPS) 23, 1009-1017, Vancouver, British Columbia, Canada, 2010. (24% acceptance rate, 1.6% acceptance rate for plenary oral presentation)
[.pdf] [.pptx] [videolecture] [poster .pdf]
The pdf is the full version of the paper, including the appendix. The printed proceedings do not include the appendix.

Jeff Johns, Sridhar Mahadevan, & Chang Wang.
Compact Spectral Bases for Value Function Approximation using Kronecker Factorization.
Proceedings of the 22nd National Conference on Artificial Intelligence (AAAI), 559-564, Vancouver, British Columbia, Canada, 2007. (27% acceptance rate)
[.pdf] [.ppt]

Jeff Johns & Sridhar Mahadevan.
Constructing Basis Functions from Directed Graphs for Value Function Approximation.
Proceedings of the 24th International Conference on Machine Learning (ICML), 385-392, Corvallis, Oregon, USA, 2007. (29% acceptance rate)
[.pdf] [.ppt]

Sridhar Mahadevan, Sarah Osentoski, Jeff Johns, Kimberly Ferguson, & Chang Wang.
Learning to Plan using Harmonic Analysis of Diffusion Models.
Proceedings of the 17th International Conference on Automated Planning and Scheduling (ICAPS), 224-231, Providence, Rhode Island, USA, 2007. (32% acceptance rate)
[.pdf]

Ivon Arroyo, Kimberly Ferguson, Jeff Johns, Toby Dragon, Hasmik Mehranian, Donald Fisher, Andrew Barto, Sridhar Mahadevan, & Beverly Woolf.
Repairing Disengagement with Non-Invasive Interventions.
Proceedings of the 13th International Conference on Artificial Intelligence in Education (AIED), 195-202, Marina Del Rey, California, USA, 2007. (30% acceptance rate)
[.pdf]

Jeff Johns & Beverly Woolf.
A Dynamic Mixture Model to Detect Student Motivation and Proficiency.
Proceedings of the 21st National Conference on Artificial Intelligence (AAAI), 163-168, Boston, Massachusetts, USA, 2006. (30% acceptance rate)
[.pdf] [.ppt]

Jeff Johns, Sridhar Mahadevan, & Beverly Woolf.
Estimating Student Proficiency using an Item Response Theory Model.
Proceedings of the 8th International Conference on Intelligent Tutoring Systems (ITS), 473-480, Jhongli, Taiwan, 2006. (32% acceptance rate)
[.pdf]
This paper appeared in the Lecture Notes in Computer Science.

Jeff Johns & Sridhar Mahadevan.
A Variational Learning Algorithm for the Abstract Hidden Markov Model.
Proceedings of the 20th National Conference on Artificial Intelligence (AAAI), 9-14, Pittsburgh, Pennsylvania, USA, 2005. (28% acceptance rate)
[.pdf]

Jeff Johns, Sarah Osentoski, & Sridhar Mahadevan.
Representation Discovery in Planning using Harmonic Analysis.
AAAI Fall Symposia on Computational Approaches to Representation Change During Learning and Development, Washington, D.C., USA, 2007.
[.pdf] [.ppt]

Anders Jonsson, Jeff Johns, Hasmik Mehranian, Ivon Arroyo, Beverly Woolf, Andrew Barto, Donald Fisher, & Sridhar Mahadevan.
Evaluating the Feasibility of Learning Student Models from Data.
AAAI Workshop on Educational Data Mining, 1-6, Pittsburgh, Pennsylvania, USA, 2005.
[.pdf]

Jeff Johns & Sridhar Mahadevan.
Sparse Approximate Policy Evaluation using Graph-based Basis Functions.
University of Massachusetts Amherst Technical Report UM-CS-2009-041, 2009.
[.pdf]

Jeff Johns, Marek Petrik, & Sridhar Mahadevan.
Hybrid Least-Squares Algorithms for Approximate Policy Evaluation.
University of Massachusetts Amherst Technical Report UM-CS-2008-045, 2008.
A more refined version of this paper appeared in the Machine Learning journal in 2009.

Proto-Value Functions
This project addresses the challenge of automatically constructing and using basis functions (proto-value functions) that are useful for solving Markov decision processes and reinforcement learning problems.

Wayang Outpost
In the Wayang Outpost project, an intelligent tutoring system for SAT-style geometry problems was developed. My role on this project was to estimate and track the dynamic behavior of a student interacting with the tutoring system.

I have been a Program Committee member for the following conferences:

I was also a Reviewer for:

Awarded a 2-year Computing Innovation Fellowship funded by the National Science Foundation (2009)

Received a National Science Foundation Research Experience for Undergraduates (NSF-REU) grant to study at Montana State University's Center for Biofilm Engineering (1997)