My primary research interest is in the area of machine learning. I am
specifically looking for ways to make ML techniques scale to large problems and
work under difficult time constraints by defining high level structure for the
organization of
knowledge representations. The Design Intelligence Laboratory led by my advisor, Prof. Ashok Goel, is also interested in model based reflection, which can be seen as similarly defining high level structure for process organization. My interests include both statistical and knowledge-based techniques, and I rely on work in both areas in my research.
I also worked with Prof. Frank Dellaert in Fall 2005/Spring 2006 on applying generalized Fourier transforms to the correspondence problem in computer vision.
Publications
Joshua Jones and Ashok Goel. "Retrospective Self-Adaptation of an Agent's
Domain Knowledge: Perceptually-Grounded Semantics for Structural Credit
Assignment". AAAI Workshop on Metareasoning. Chicago, IL, 2008. pdf
Joshua Jones and Ashok Goel. "Structural Credit Assignment in Hierarchical
Classification". ICAI-07. Las Vegas, NV, 2007. pdf
Joshua Jones, Ashok Goel, and Spencer Rugaber. "Teleological Modeling and Reasoning for Automated Software Adaptations". DESRIST-07. Pasadena, CA, 2007. pdf
Joshua Jones and Ashok Goel. "Knowledge Organization and Structural Credit Assignment". IJCAI Workshop on Reasoning, Representation and Learning in Computer Games. Edinburgh, UK, 2005. pdf
Patrick Ulam, Ashok Goel, Joshua Jones and William Murdoch. "Using Model-Based Reflection to Guide Reinforcement Learning". IJCAI Workshop on Reasoning, Representation and Learning in Computer Games. Edinburgh, UK, 2005. pdf
Joshua Jones and Ashok Goel. "Hierarchical Judgement Composition: Revisiting the Structural Credit Assignment Problem". AAAI Workshop on Challanges in Game AI . San Jose, CA, 2004. pdf
Patrick Ulam, Ashok Goel, and Joshua Jones. "Reflection in Action: Model-Based Self-Adaptation in Game Playing Agents". AAAI Workshop on Challanges in Game AI . San Jose, CA, 2004. pdf
PPServer
Oft requested info about getting set up with the Parallel Problem Server
(geared to GT students, but may be more generally useful as a starting point)
can be found
here. I worked
on code for the PPServer with Prof. Charles Isbell during a past project.
7001 Projects
Three small, short term projects completed during my first semester at Georgia
Tech as part of the introductory course for incoming PhD students.
