(Georgia Institute of Technology, 2005)
Oh, Sang Min; Rehg, James M.; Dellaert, Frank
We introduce Segmental Switching Linear Dynamic Systems (S-SLDS), which improve on standard
SLDSs by explicitly incorporating duration modeling capabilities. We show that S-SLDSs can adopt
arbitrary finite-sized duration models that describe data more accurately than the geometric distributions
induced by standard SLDSs. We also show that we can convert an S-SLDS to an equivalent standard
SLDS with sparse structure in the resulting transition matrix. This insight makes it possible to adopt
existing inference and learning algorithms for the standard SLDS models to the S-SLDS framework. As
a consequence, the more powerful S-SLDS model can be adopted with only modest additional effort in
most cases where an SLDS model can be applied. The experimental results on honeybee dance decoding
tasks demonstrate the robust inference capabilities of the proposed S-SLDS model.
(Georgia Institute of Technology, 2002)
Wolenetz, Matthew David; Mandviwala, Hasnain A.; Adhikari, Sameer; Angelov, Yavor; Ramachandran, Umakishore; Mackenzie, Kenneth M.; Rehg, James M.
Interactive multimedia applications (such as audio/video processing) are
good candidates for cluster computing. Such applications are best represented
as coarse-grain dataflow graphs and are rich in pipelined, task, and data
parallelism. Specification of the strategies for mapping computational abstractions to compute nodes, and their plumbing are two
important issues in the design of such complex parallel and distributed
applications. Due to the varieties of parallelism that are available in such
applications, the space of strategies to be explored can be vast. We have
developed an aspect-oriented programming language for cluster computing called STAGES. This language allows the algorithm design
to be disentangled from the connection management and performance concerns.
STAGES provides a simple syntax for specifying the connections among threads
and data abstractions, and their mapping onto the nodes of the cluster. The
current implementation targets the Stampede cluster programming library.
However, the language is general and can be retargeted to a different set of
abstractions.
In this paper, we present STAGES, its implementation, and its utility for
mapping complex applications onto a cluster. We also present performance
results from exploring the parallelism space for two such applications on a
17-node cluster of 8-way SMPs (Intel Xeon processors) interconnected by
Gigabit Ethernet.