Garrett Ethan Katz
Electrical Engineering & Computer Science
- 3-125 CST
- [email protected]
- B.A. Philosophy, Cornell University, 2007
- M.A. Mathematics, City College of New York, 2011
- Ph.D. Computer Science, University of Maryland, College Park, 2017
- Neural Computation
- Cognitive Robotics
- Dynamical Systems
My current research focuses on programmable neural networks: neural networks that can be “programmed” like a conventional computer to execute symbolic, cognitive-level tasks, but can then refine that procedural knowledge by learning from examples and experience. One application of this work is in robotic imitation learning: “programming” robots from a single human demonstration of a task that requires high-level planning and reasoning. A second application of this work is modeling the neural basis of cognition, and cognitive disorders, in humans.
In other research I am developing new solution methods for fixed point location in recurrent neural networks and other dynamical systems, including gradient flows of optimization objective functions. This work applies broadly to solving non-linear systems of equations and non-convex optimization. I have also worked on methods for computational tomography of biological virus particles.
My teaching interests include machine learning and artificial intelligence, especially neural computation and automated planning, as well as dynamical systems, robotics, and human-robot interaction.
- Best student paper award at the 9thInternational Conference on Artifical General Intelligence, 2016
- Distinguished Graduate Student Teacher, University of Maryland, 2014
- Katz GE, Reggia JA (2017). Using Directional Fibers to Locate Fixed Points of Recurrent Neural Networks. IEEE Transactions on Neural Networks and Learning Systems. DOI 10.1109/TNNLS.2017.2733544
- Katz GE, Huang DW, Hauge T, Gentili RJ, Reggia JA (2017). A Novel Parsimonious Cause-Effect Reasoning Algorithm for Robot Imitation and Plan Recognition. IEEE Transactions on Cognitive and Developmental Systems. IEEE. DOI 10.1109/TCDS.2017.2651643