[Ml-stat-talks] Thursday, August 12, 2010 at 10:30 am: "Stochastic Search, Optimization and Regression with Energy Applications" (Lauren Hannah)

David Blei blei at CS.Princeton.EDU
Mon Aug 9 22:45:04 EDT 2010

hi ml-stat-talks

princeton's own lauren hannah will be giving a talk on "stochastic search,
optimization and regression with energy applications."  this is her FPO.
 lauren will be presenting an impressive set of research results, both
theoretical and empirical.  please come if you are interested in any of:

 (a) bayesian nonparametric methods
 (b) optimization, stochastic optimization, and decision making
 (c) applications to energy and the environment

details are below.  i hope to see you there!



"Stochastic Search, Optimization and Regression with Energy Applications"
Lauren Hannah, Princeton University (soon to be Duke University)
Thursday August 12, 10:30AM
Room 125, Sherrerd Hall

Designing clean energy systems will be an important task over the next few
decades.  One of the major roadblocks is a lack of mathematical tools to
economically evaluate those energy systems.  However, solutions to these
mathematical problems are also of interest to the operations research and
statistical communities in general.  I will discuss two problems that are of
important to the energy community itself or provide support for solution
methods used in energy problems:  Bayesian nonparametric regression and
stochastic search with an observable state variable.

First, I discuss Dirichlet Process mixtures of Generalized Linear Models
(DP-GLM), a new method of nonparametric regression that accommodates
continuous and categorical inputs, and responses that can be modeled by a
generalized linear model.  I give conditions for the asymptotic unbiasedness
of the DP-GLM regression mean function estimate and examples for when those
conditions hold.  I empirically examine the properties of the DP-GLM and
evaluate it on several datasets.

Second, I discuss convex stochastic search problems with an observable state
variable.  Currently, there is no general purpose algorithm to solve this
class of problems.  I propose using nonparametric density estimation to take
observations from the joint state-outcome distribution and infer the optimal
decision for a given query state.  I give two solution methods that depend
on the problem characteristics: function-based and gradient-based
optimization.  I examine two weighting schemes, kernel-based weights and
Dirichlet process-based weights, for use with the solution methods.  The
weights and solution methods are tested on a synthetic multi-product
newsvendor problem and the hour-ahead wind commitment problem.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20100809/03f6bd56/attachment.html>

More information about the Ml-stat-talks mailing list