Embedded Files
Post date: Mar 18, 2012 1:21:04 PM
Mr. Francis George C. Cabarle will be defending his Masters thesis entitled:
"Spiking Neural P Systems: Implementations and Applications"
Abstract:
In this work the class of membrane systems under Membrane Computing which are inspired by the spiking of biological neurons, known as Spiking Neural P systems (simply, SNP systems), are investigated. SNP systems, similar to other P system variants, are parallel, nondeterministic, and distributed computing devices. An algorithm which simulates the computations of SNP systems is first presented. SNP systems are then implemented and simulated sequentially in the CPU. From the sequential simulator, a parallel simulator is developed using graphics processing units (simply, GPUs). The parallel simulator implements the nondeterministic computations of SNP systems in a breadth-first manner. GPUs manufactured by NVIDIA are used, since the parallel and distributed computations of SNP systems match the massively parallel architecture of GPUs. Benchmark SNP systems, which are connected graphs, are developed to stress test the parallel simulator. For a 16 neuron benchmark SNP system with $65,536$ nondeterministic rule selection choices, the speedup of the parallel simulator relative to the sequential simulator is shown to be up to 2.31 times.
Additionally, structural and behavioral relationships of SNP systems and Petri nets are elaborated. Petri nets are mathematical and graphical tools for modeling and anaylsis. It is shown that Petri nets and their fundamental routing constructs can be translated into SNP systems and vice-versa. These fundamental routing constructs include parallel and choice routing, relating Petri net structure to SNP systems. Behavioral properties of Petri nets such as liveness and boundedness are translated and related to SNP systems as well. The translated structural and behavioral properties of SNP systems are then used to model and verify correct workflow processes.
Keywords: Natural Computing, Membrane Computing, Spiking Neural P systems,Parallel Computing, GPU Computing, CUDA, Petri Nets
Adviser:
Henry N. Adorna, Ph.D.
Panel Members:
Pablo Manalastas, Ph.D.
Jaime D.L. Caro, Ph.D.
Joy Reyes-Madamba
Cedric Festin, Ph.D.
Date: 26 March 2012
Time: 10AM – 12PM
Venue: ERDT room, Department of Computer Science,
Velasquez St., UP Diliman, Quezon city
Faculty, students, and the general public are invited.
and
Ms. Richelle Anne B. Juayong will be defending her Masters thesis entitled:
"Computing in Evolution-Communication P systems with Energy"
Abstract:
Due to the limitations of today’s standard computers, the natural computing paradigm introduces biologically-inspired computational models and techniques in order to device new ways of computing. A relatively new branch of the natural computing paradigm called membrane computing abstracts features and functions occurring onliving cells. The theoretical model used for membrane computing is called a P system. In this work, we are particularly interested in a specific variant of P system called Evolution-Communication P system with energy (ECPe system) which was introduced in 2009 to initiate communication complexity in the domain of P systems. We take part in the analysis of ECPe system’s computation by exploring its computational power and matrix representation. Specifically, we improve previous universality results of ECPe systems by incorporating rules that are similar to the antiport process of transporting molecules in cells. We also investigate the use of dynamical communication measures derived from computations on ECPe systems in examining the quality of communication through so-called Sevilla carpets. Finally, we represent computations in ECPe systems as matrix operations and answer theproblem of computing forward and backward by extending previous works in other P system variants. Our motivation is the use of algebraic representation in analyzing computations in ECPe systems as proven in other types of P system and the possible simulation of ECPe systems computations in parallel processors, specifically, in Graphical Processing Units (GPUs).
Adviser:
Henry N. Adorna, Ph.D.
Panel Members:
Pablo Manalastas, Ph.D.
Jaime D.L. Caro, Ph.D.
John Paul Vergara, Ph.D.
Adrian Valdez, Ph.D.
Date: 26 March 2012
Time: 3PM – 5PM
Venue: ERDT room, Department of Computer Science,
Velasquez St., UP Diliman, Quezon city
Faculty, students, and the general public are invited.
Page updated
Google Sites
Report abuse