Greg Wolffe

Memory-efficient implementation of a graphics processor-based cluster detection algorithm for large spatial databases

Numerous approaches have been proposed for detecting clusters, groups of data in spatial databases. Of these, the algorithm known as Density Based Spatial Clustering of Applications with Noise (DBSCAN) is a recent approach which has proven efficient for larger databases. Graphical Processing Units (GPUs), used originally to aid in the processing of high intensity graphics, have been found to be highly effective as general purpose parallel computing platforms. In this project, a GPU-based DBSCAN program has been implemented: the enhancement in this program allows for better memory scalability for use with very large databases. Algorithm performance, as compared to the original sequential program and to an initial GPU implementation, is investigated and analyzed.

Parallel implementations of FGMRES for solving large, sparse non-symmetric linear systems

Abstract The Flexible Generalized Minimal Residual method (FGMRES) is an attractive iterative solver for non-symmetric systems of linear equations. This paper presents several methods for parallelizing FGMRES for a variety of archi- tectures including multi-core CPU, Graphics Processing Units (GPU), and multi-GPU systems. The parallel imple- mentations utilize OpenMP and CUDA kernels, and are organized according to thread scope. The linear systems employed in this study correspond to the discrete analogues of realistic three-dimensional convection-diffusion problems, and range in size to nearly 107 linear equations. All of the parallel implementations, particularly the novel hybrid approach, show a significant speedup over the sequential version. Theoretical insight and perfor- mance data is provided to allow informed decisions as to the most effective parallelization method for a given architecture.

PyGASP: Python-based GPU-accelerated signal processing

Computational science is the application of computing technology to evaluate mathematical models in order to solve problems in the scientific disciplines. Many scientific fields are experiencing an explosion of data, with signal processing being a crucial technique for aiding interpretation and for distinguishing meaningful information from noise. This process requires tools that can be easily used by researchers from all branches of science and which are fast enough to manage the enormous amount of data being generated. This paper presents such a toolkit: an intuitive, high-performance Python library for facilitating large-scale signal analysis. Of particular interest is a novel PyCUDA implementation of the Discrete Wavelet Transform (DWT), several applications of which are demonstrated in this paper.

Query Expansion in Enterprise Search

Although web search remains an active research area, interest in enterprise search has not kept up with the information requirements of the contemporary workforce. To address these issues, this research aims to develop, implement, and study the query expansion techniques most effective at improving relevancy in enterprise search. The case-study instrument was a custom Apache Solr-based search application deployed at a medium-sized manufacturing company. It was hypothesized that a composition of techniques tailored to enterprise content and information needs would prove effective in increasing relevancy evaluation scores. Query expansion techniques leveraging entity recognition, alphanumeric term identification, and intent classification were implemented and studied using real enterprise content and query logs. They were evaluated against a set of test queries derived from relevance survey results using standard relevancy metrics such as normalized discounted cumulative gain (nDCG). Each of these modules produced meaningful and statistically significant improvements in relevancy.

Proximity synchronization for mobile wireless sensor networks

Wireless sensor networks are designed to be used in any situation that requires monitoring of widely-dispersed geographic areas. They are capable of providing automated monitoring with high precision at low cost over long periods of time. In order to achieve the conflicting goals of high precision at low cost, software techniques can be employed to enhance the precision of important metrics. In particular, this investigation introduces an improved protocol for synchronizing the clocks of different sensor nodes, for the purpose of improving the accuracy of time-sensitive data acquisition. In the course of our initial investigation, several weaknesses and deficiencies exhibited by existing synchronization protocols were identified, particularly with respect to mobile sensor networks. To address these issues a new protocol was developed. The protocol is based on the Network-wide Time Synchronization in sensor networks protocol, modified to use a restricted tree-based hierarchy arranged as a minimum spanning tree. A key feature of the improved protocol is that it distributes the work of synchronization, thereby eliminating potential bottleneck constraints. In extensive testing, it was also shown to reduce the number of message collisions to which larger networks are prone. When compared to the original protocol, these improvements resulted in lower clock synchronization error across the network and substantially reduced sensor power consumption. In addition, the improved protocol is more robust, in that it has the ability to dynamically reconnect and resynchronize as mobile nodes move in and out of communication range.