Machine Learning and AI Applications in Astrophysics and Cosmology

A workshop associated with the IJCAI-09 conference, http://ijcai-09.org/


Workshop co-organizers:

Prof. S. George Djorgovski, Caltech, george*astro.caltech.edu
Prof. Alexander Gray, Georgia Tech, agray*cc.gatech.edu
Prof. Alex Szalay, JHU, szalay*jhu.edu
[*=@]

Astronomy has become an enormously data-rich science, with digital sky
surveys detecting literally billions of sources across a full range of
the electromagnetic spectrum, and measuring hundreds attributes for each.
Such surveys and other data sets are now being federated within the Virtual
Observatory framework.

An effective data exploration and knowledge discovery from such highly complex
and massive sets poses significant challenges, and calls for a more extensive
use of AI/ML techniques, which should become a standard part of a scientist's
toolkit.  At the same time, interesting, challenging, but doable problems
originating from the astrophysics and cosmology could help drive progress
in the AI/ML field, and lead to broader advances.  There is a great synergistic
opportunity here, for both of these communities.

This workshop will bring together the pioneering or established practitioners
of the use of ML/AI tools and methodologies in astronomy and cosmology, as well
as some newcomers to the field.  We will review briefly the past accomplishments
and current practices and challenges, but the main focus would be on defining
the new paths for future exploration and collaboration, including some specific
problems and challenges which could be tackled over the next few years.  We will
aim for a roughly equal balance between computing-savvy astrophysicists and
applied computer scientists.

Our goal is to foster stronger connections between the AI/ML community and the
data intensive astronomy / virtual observatory community.  We will aim for a
roughly equal balance between astronomers and computer scientists, an for a
total number of about 30 to 40 participants, small enough for some real
discussions, and yet achieving a critical mass.

We aim to inform better both of these communities about each other's interests
and challenges, and to facilitate and foster new interdisciplinary collaborations
between them.  This meeting will also serve as a forum to evaluate critically
the past successes and failures, and derive some constructive lessons from them.
Finally, we expect that the participants would go on to become even more effective
community evangelists for the use of AI/ML methods in data-intensive astronomy
and cosmology.

The workshop would last 2 days, and would be heavily biased towards discussions,
both structured/moderated and informal (during the breaks).  We will invite a
number of keynote or review talks, with explicit instructions to the speakers
to focus on the challenges and prospects for the future, and to be constructively
provocative, rather than to simply describe their recent results.  There would be
no contributed talks, but we would allow for posters.