Thông tin chung

  English

  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác

  Tài liệu tham khảo

  Hiệu chỉnh

 
Số người truy cập: 106,068,190

 Evaluation of Artificial Neural Network Architectures for Pattern Recognition on FPGA
Tác giả hoặc Nhóm tác giả: Thang Viet Huynh
Nơi đăng: International Journal of Computing and Digital Systems (ISSN: 2210-142X) DOI: http://dx.doi.org/10.12785/ijcds/060305; Số: Volume 6 - Issue 3;Từ->đến trang: 133-138;Năm: 2017
Lĩnh vực: Khoa học công nghệ; Loại: Bài báo khoa học; Thể loại: Quốc tế
TÓM TẮT
In this paper, we present the design and implementation of two hardware architectures, namely MHL-ANN and SHL-ANN, for the realization of artificial neural networks on reconfigurable computing platforms like FPGA. We use 16-bit half-precision floating-point number format to represent the weights of the designed networks. The networks are synthesized and verified on Xilinx Virtex-5 XC5VLX-110T FPGA. We study the scalability and hardware resource utilization of the two proposed neural network architectures. For performance evaluation, the handwritten digit recognition application with MNIST database is performed, which reported a recognition rate of 90.88% when using an MHL-ANN architecture of size 20-12-10 and a recognition rate of 96.83% when using an SHL-ANN architecture of size 784-40-10. Experimental results showed that the SHL-ANN architecture is very potential for high performance embedded recognition applications.
ABSTRACT
In this paper, we present the design and implementation of two hardware architectures, namely MHL-ANN and SHL-ANN, for the realization of artificial neural networks on reconfigurable computing platforms like FPGA. We use 16-bit half-precision floating-point number format to represent the weights of the designed networks. The networks are synthesized and verified on Xilinx Virtex-5 XC5VLX-110T FPGA. We study the scalability and hardware resource utilization of the two proposed neural network architectures. For performance evaluation, the handwritten digit recognition application with MNIST database is performed, which reported a recognition rate of 90.88% when using an MHL-ANN architecture of size 20-12-10 and a recognition rate of 96.83% when using an SHL-ANN architecture of size 784-40-10. Experimental results showed that the SHL-ANN architecture is very potential for high performance embedded recognition applications.
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn