JOB: Postdoc at UMBC in AI planning/learning

The MAPLE [1] and CORAL [2] labs at the University of Maryland
Baltimore
County are seeking a Postdoctoral Associate to work on a new project
funded
by the DARPA Integrated Learning program. The goal of that program is
to
build systems that learn general plans or processes from users by being
shown one example. Research at UMBC will involve the following:

- developing methods by which a goal-driven learning system can
determine
uncertainty estimates and decision criteria for incorporation into
hierarchical task plan models

- developing methods by which a learning system can learn task
structure
models equivalent to context-free grammars from observations
coupled
with user advice and additional experimental evidence generated by
internal experiments

For more information, contact either Prof. Marie desJardins
(mariedj@xxxxxxxxxxx) or Prof Tim Oates (oates@xxxxxxxxxxx). The
appointment will be for one year, with continuing support contingent on
continuing funding from DARPA. Salary is negotiable.

[1] Multi-Agent Planning and Learning - http://maple.cs.umbc.edu
[2] Cognition, Robotics, and Learning - http://www.coral-lab.org

[ comp.ai is moderated. To submit, just post and be patient, or if ]
[ that fails mail your article to <comp-ai@xxxxxxxxxxxxxxxxxx>, and ]
[ ask your news administrator to fix the problems with your system. ]
.

Relevant Pages

  • Re: Innovation in Humans
    ... You cannot build a generic learning system, ... When you build a generic reinforcement learning system, ... not to maximize only _current_ reward values. ... (assuming the system is strong enough to find those behaviors). ...
    (comp.ai.philosophy)
  • Real AI
    ... don't try and "explain" it again my brain might ... from learning anything interesting. ... completely consistent with classical and operant conditioning. ... have to figure out what type of learning system are needed to ...
    (comp.ai.philosophy)
  • Re: Can operant conditioning account for all learning?
    ... And one thing the process of evolution is known for, ... Even though we are adaptive learning machine, ... evolution is itself a slow learning system which tends to build ... is that we don't have learning hardware to match any ...
    (comp.ai.philosophy)
  • Re: Diagrams of learning was Re: self-building intelligence
    ... at the levels of sorting individual pulses. ... "What you need to do instead, is create a learning system that has the ... If so an indication of how the system builds better learning processes ...
    (comp.ai.philosophy)
  • Re: the Fish and the Robot
    ... simplistic learning protocol might produce a result where only a single ... I don't see any reason to believe a learning system will do this on its ... your temporal stage-sequencing "smart overseer". ... one move in the middle of the game might be better than another. ...
    (comp.robotics.misc)