6-layer convolutional neural network  (SMAC Corp)

Journal: PLoS ONE

Article Title: Research on OpenCL optimization for FPGA deep learning application

doi: 10.1371/journal.pone.0222984

Figure Lengend Snippet: Parameters of a single convolution layer.

Article Snippet: Moreover, this paper also tests the performance of the convolution layer optimization program provided by Xilinx, as summarized in where it can be seen that the final performance of the program is 29 times higher than that of the optimization program provided by Xilinx company.

Techniques:

Journal: PLoS ONE

Article Title: Research on OpenCL optimization for FPGA deep learning application

doi: 10.1371/journal.pone.0222984

Figure Lengend Snippet: The specific code optimization of convolution multiplication calculation.

Article Snippet: Moreover, this paper also tests the performance of the convolution layer optimization program provided by Xilinx, as summarized in where it can be seen that the final performance of the program is 29 times higher than that of the optimization program provided by Xilinx company.

Techniques:

Journal: PLoS ONE

Article Title: Research on OpenCL optimization for FPGA deep learning application

doi: 10.1371/journal.pone.0222984

Figure Lengend Snippet: Performance comparison of the different optimization programs in the convolution layer OpenCL.

Article Snippet: Moreover, this paper also tests the performance of the convolution layer optimization program provided by Xilinx, as summarized in where it can be seen that the final performance of the program is 29 times higher than that of the optimization program provided by Xilinx company.

Techniques: Comparison

Journal: PLoS ONE

Article Title: Research on OpenCL optimization for FPGA deep learning application

doi: 10.1371/journal.pone.0222984

Figure Lengend Snippet: Performance comparison of different convolution scale optimization program with Xilinx company.

Article Snippet: Moreover, this paper also tests the performance of the convolution layer optimization program provided by Xilinx, as summarized in where it can be seen that the final performance of the program is 29 times higher than that of the optimization program provided by Xilinx company.

Techniques: Comparison

Journal: PLoS ONE

Article Title: Research on OpenCL optimization for FPGA deep learning application

doi: 10.1371/journal.pone.0222984

Figure Lengend Snippet: Optimized speed-up ratio of different scale convolution.

Article Snippet: Moreover, this paper also tests the performance of the convolution layer optimization program provided by Xilinx, as summarized in where it can be seen that the final performance of the program is 29 times higher than that of the optimization program provided by Xilinx company.

Techniques: