Keynote Speaker:
    Pat Langley, Institute for the Study of Learning and Expertise

 Knowledge, Data, and Search in Computational Discovery (ppt slides)

                             Pat Langley
                  Computational Learning Laboratory
           Center for the Study of Language and Information
                         Stanford University
                  http://cll.stanford.edu/~langley/


Early research on machine learning, which had strong links to symbolic
artificial intelligence, studied interactions among three factors:
knowledge, data, and search. Over the past decade, machine learning
and statistics have joined forces to develop powerful techniques that
combine data and search but that disregard the role of knowledge. In
this talk, I argue that computational learning and discovery systems
would benefit from a return to explicit, symbolic representations of
knowledge in both their inputs and their outputs. I illustrate this
approach with some recent results on the construction and revision
of scientific models that are cast as sets of explanatory processes.
In closing, I outline some open research problems in machine learning
and discovery that revolve around the reintegration of knowledge with
data and search.


This talk describes joint work with Nima Asgharbeygi, Will Bridewell,
Andrew Pohorille, Oren Shiran, Jeff Shrager, and Ljupco Todorovski.