Co-developed with Yelp,
this package performs parallel global optimization
of expensive functions. It was developed for optimizing website metrics
like click-through rate, but can also be used for non-web applications.
This Matlab library implements independent KG from (Frazier, Powell & Dayanik, in SIAM J. Control & Optimization 2008) and correlated KG from (Frazier, Powell & Dayanik, in INFORMS J. on Computing 2009).
The InfoCollection library has java implementations for a number of ranking and selection policies in the normal sampling case with both variance known and unknown. I intend upon expanding it to simulate a larger class of information class problems.
4/9/2008 (jar), (source).
This version of the library requires the "Miscellaneous Utilies" library below as well as the
Additionally, a few of the classes also require a numerical integration package, flanagan.jar, or a small library for handling tuples from adventuresinsoftware.com.
This java library contains generic utilities that are not specific to information collection. It is used by the InfoCollection library.
(jar), (source) (updated 4/9/2008).
InfoCollection-Based Numerical Experiments
This code was used directly to generate published numerical results, and generally uses the InfoCollection library.
It also uses gnuplot to generate plots.
Winter Simulation Conference 2008: source. This code uses the 4/9/2008 version of the InfoCollection library.
Winter Simulation Conference 2009 (with Stephen Chick): source. This code was updated 07/2013 to make it easier to use. It uses the 12/11/2009 version of the InfoCollection library, which is included in the zip file.
The stepsize library is a collection of java files implementing various stepsize rules for approximate dynamic programming (Library version updated 4/26/2007).
This collection of matlab files related to timing uncertainty contains an implementation of a dynamic programming algorithm for computing the optimal policy for doing sequential hypothesis testing under a stochastic deadline. This software was put together in support of a NIPS conference paper that Angela Yu and I wrote in 2007. The archive contains a README describing the files. (updated 8/17/2007).