Improving the efficiency of gabriel graph-based classifiers for hardware-optimized implementations
Carregando...
Data
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Federal de Minas Gerais
Descrição
Tipo
Artigo de evento
Título alternativo
Primeiro orientador
Membros da banca
Resumo
This work evaluates strategies to reduce the computational cost of Gabriel graph-based classifiers, in order to make them more suitable for hardware implementation. An analysis of the impact of the bit precision provides insights on the model's robustness for lower precision arithmetic. Additionally, a parallelization technique is proposed in order to improve the efficiency of the support edges computation. The results show that the lower bit precision models are statistically equivalent to the reference double-precision ones. Also, the implementation of the proposed parallel algorithm provides a significant reduction in the running time when applied in large datasets, while maintaining its accuracy.
Abstract
Assunto
Computadores (Engenharia)
Palavras-chave
Hardware , Field programmable gate arrays , Computational modeling , Numerical models , Data models , Analytical models , Parallel processing, Classification , Gabriel graph , FPGA , numerical representation , parallel computing, Running Time , Analysis Of The Impact , Precise Model , Low Precision , Hardware Implementation , Least Significant Bit , Parallel Algorithm , Double Precision , Parallel Technique , Machine Learning , Model Performance , Training Data , Exponential Growth , Average Results , Graphics Processing Unit , Class Labels , Resource Consumption , Hyperplane , Reference Model , Graph Structure , Hardware Architecture , Vertex Degree , Single Precision , Numerical Representation , Filtering Techniques
Citação
Curso
Endereço externo
https://ieeexplore.ieee.org/document/8730227