Table 4 The average execution time comparison between the MaxwellFDFD simulator and the CNN model.

From: Simulation acceleration for transmittance of electromagnetic waves in 2D slit arrays using deep learning

# of samples

MaxwellFDFD

CNN

GPU (s)

Non-GPU (s)

1

60 s

0.0004

0.0013

9,000

150 h

3.3

11.3

27,000

450 h

10

34

  1. The execution time was calculated during the Training set (27,000 samples) and the Test set (9,000 samples). CNN model was approximately 160,000 times faster than the simulator when using the graphics processing unit (GPU) and 48,000 times faster without GPU.
Back to article page