Banca de QUALIFICAÇÃO: LEONARDO ALVES DIAS

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE : LEONARDO ALVES DIAS
DATA : 14/12/2018
HORA: 13:30
LOCAL: Sala 2 DCA
TÍTULO:

Proposal of Parallel Implementation of Clustering Algorithms with Reconfigurable Computing

PALAVRAS-CHAVES:

Parallel system, FPGA, data clustering, reconfigurable computing.

PÁGINAS: 65
GRANDE ÁREA: Engenharias
ÁREA: Engenharia Elétrica
SUBÁREA: Circuitos Elétricos, Magnéticos e Eletrônicos
ESPECIALIDADE: Circuitos Eletrônicos
RESUMO:

This work presents a study of data clustering algorithms implemented on reconfigurable hardware for general applications, in order to increase the speed of data processing. Clustering algorithms have been widely applied to find the correlation between data in different areas. However, these algorithms usually imply a high processing complexity and, in addition, the amount of data that is currently stored is very large. Thus, the need for high-speed data processing to perform its analysis has become even more critical, especially for real-time applications. One solution that was adopted to increase processing speed is the use of parallel techniques implemented in Field-Programmable Gate Arrays (FPGAs), which proved to be more efficient compared to sequential systems. Therefore, this work proposes the completely parallel implementation of data clustering algorithms in FPGA to optimize the processing time of the systems in several areas, allowing applications for systems with massive amounts of data. A new proposal for the implementation of the K-means clustering algorithm is presented, along with analyzes of the results related to the processing time (or throughput) and the FPGA area occupation (or hardware resource) for different parameters, reaching processing rates more than 53 million data per second. Comparisons with state of the art are also presented, showing speedups greater than 15573x. The implementation presented here points to a new direction associated with the implementation of clustering algorithms and can be used in other algorithms.

MEMBROS DA BANCA:
Presidente - 1837240 - MARCELO AUGUSTO COSTA FERNANDES
Interno - 2885532 - IVANOVITCH MEDEIROS DANTAS DA SILVA
Externo ao Programa - 1669545 - DANIEL SABINO AMORIM DE ARAUJO
Externo à Instituição - JOÃO PAULO DE CASTRO CANAS FERREIRA - FEUP