[Ml-stat-talks] Karsten Borgwardt: Scalable Graph Kernels

David Blei blei at CS.Princeton.EDU
Fri Jul 2 10:20:53 EDT 2010

hi ml-stat-talks,

karsten borgwardt will be speaking next wednesday at 11:30AM in CS
room 402.  the talk is on "scalable graph kernels" and looks to be
very interesting.  hope to see you there!

if you'd like to chat with karsten, please email him directly.  he's
around all day---though he may be busy around 2:30pm depending on the
outcome of today's soccer match.



Computer Science Building 402

Dr. Karsten Borgwardt,
Machine Learning and Computational Biology Research Group,
Max Planck Institutes Tübingen, Germany

Scalable graph kernels

Kernel methods are a family of algorithms in intelligent data analysis,
which have gained enormous popularity in machine learning over the
last 15 years. One reason for their attractiveness
lies in the fact that the underlying theory of these algorithms can
easily be generalized from
vectorial data to strings, time series, graphs and other types of
structured data.
In real-world applications, the efficient computation of the kernel
function, i.e. of the similarity measure,
is a key challenge. For strings and time series, efficient computation
techniques for kernels were developed early on,
but graph kernels remained slow to compute and only applicable to
graphs with a few dozen nodes without attributes. A major focus of my
has been to turn graph kernels from a theoretical concept into a
useful tool for practical graph data analysis.
In my talk, I will present my work on efficient graph kernels, in
particular a recent breakthrough from 2009, which now allows for
highly scalable graph kernel computation (Shervashidze, Borgwardt.
Fast Subtree Kernels on Graphs, NIPS  2009).


More information about the Ml-stat-talks mailing list