Posted to my e-mail address:
The Journal of Machine Learning Research (www.jmlr.org) is pleased to
announce the publication of a new paper:
---------------------------------------------------------------------------------
Model Selection Through Sparse Maximum Likelihood Estimation for
Multivariate Gaussian or Binary Data
Onureena Banerjee, Laurent El Ghaoui, Alexandre d'Aspremont;
JMLR 9(Mar):485--516, 2008.
Link: http://www.jmlr.org/papers/volume9/banerjee08a/banerjee08a.pdf
Abstract
We consider the problem of estimating the parameters of a Gaussian or
binary distribution in such a way that the resulting undirected
graphical model is sparse. Our approach is to solve a maximum likelihood
problem with an added l1-norm penalty term. The problem as formulated is
convex but the memory requirements and complexity of existing interior
point methods are prohibitive for problems with more than tens of nodes.
We present two new algorithms for solving problems with at least a
thousand nodes in the Gaussian case. Our first algorithm uses block
coordinate descent, and can be interpreted as recursive l1-norm
penalized regression. Our second algorithm, based on Nesterov's first
order method, yields a complexity estimate with a better dependence on
problem size than existing interior point methods. Using a log
determinant relaxation of the log partition function (Wainwright and
Jordan, 2006), we show that these same algorithms can be used to solve
an approximate sparse maximum likelihood problem for the binary case. We
test our algorithms on synthetic data, as well as on gene expression and
senate voting records data.
---------------------------------------------------------------------------------
This paper and previous papers are available electronically at
http://www.jmlr.org
in PDF format. The papers of Volumes 1-4 were also
published in hardcopy by MIT Press; please see
http://mitpress.mit.edu/JMLR
for details. Volume 5 and subsequent
volumes are being printed in hardcopy by Microtome Publi****ng. Please
see http://www.mtome.com/Publications/JMLR/jmlr.html
for details and
ordering information.
Richard Maclin
rmaclin@[EMAIL PROTECTED]
mailing list
Jmlr-announce@[EMAIL PROTECTED]
exceeds the limit of variance in The Language of Universal
Translation--Annotated in truisms and falsity for an untimely account of
entanglement since the beginning. Goes on to reference citations that
could
not be proved as the load balancing proxy seems to be their objection as
for
making aquisitions on queer matter really (not) instead of a publishers
prerogative--a service provider seed for the monopolistic owner****p of all
information? Adding extra 'sub-linear' checks to data-sycles of
para-normal
phenomonon in the design and run-time cycles, as they've built here (since
the original) Why this was changed so out of scope as if actually to
remove
the far more interesting entanglement of congruency and crystallization
that
these network diagrams anhillate by sheer volume of any realm of the
Usenet,
supposedly, or usery of established relation****ps for which derivatives
would be implicated in memetic weapons registration. Criticism from this
commuty engage a rationalization of dynamic data exchange attributed to
the
actual agency of register independence, exclaims this editor , "what the
fat
felt".
What a periodic joke. All I can say except, what a good ECC inside of me!
Discrete Systems in Non-Linear Space for dynamic copyrights on thus, the
single life-form responsible for this 'frame' of reference. I seek a
greater characerizaion of such techniques, not a characteristic strength
for
which these 'networks' are corruptible. Trying to prove tele****tation
beyond the modal data-rate is not only infeasible, it is a self-directed
behaviour, not an adaptation.
PAL,
Patrick Ashley Meuser"-Bianca"
Cyberneticist
http://www.usag-ac.info
|
