Research Abstracts Online
January 2009 - March 2010
St. Cloud State University
College of Science and Engineering
Department of Computer Science
PI: Ji Hu Meichsner
Parallel Implementation for ICA Algorithms
Distributed Localization of Wireless Sensor Networks Using Self-organizing Maps
Implementation of Asynchronous Iterative Algorithms
These researchers have been involved in three projects using the supercomputers. The first deals with independent component analysis (ICA), which was originally developed to deal with blind signal separation problems. Recent interest in ICA has resulted in the development of interesting applications, such as electrical recordings of brain activity as given by electroencephalogram, feature analysis, and image analysis; this means that a faster algorithm for ICA becomes important. The goal of this project is to implement such an algorithm on a parallel machine in order to improve its efficiency.
In the second project, begun during this period, the researchers are developing a decentralized localization method for wireless sensor networks, based on self-organizing maps. As larger sets of wireless sensor networks are being deployed, an important characteristic of the network that could enhance its capabilities is position awareness. While several approaches have been proposed for localization, that is, position awareness without using the Global Positioning System, most techniques are either centralized or rely on anchor nodes. In this project, the algorithm is implemented for different size networks and the simulation results show the algorithm is efficient when compared to single processor or centralized localization methods; further the approach does not require anchor nodes. An error analysis shows that the proposed approach is a feasible method for computing the localization of sensor networks using a distributed architecture.
The third project, completed during this period, investigated how to effectively implement the asynchronous parallel iterative algorithms on message-passing systems using the MPI non-blocking communication model. The main idea was to use the MPI_IPROBE function to check for the existence of pending messages without receiving them, thereby allowing the researchers to write programs that interleave local computation with the processing of incoming messages.
Swarada Navele, Graduate Student
Bishow Paudel, Graduate Student