Stopping criteria. We already have shown the convergence properties of our MG algorithm, and now we clarify the stopping criteria. Although R and RR defined above can be used to stop the algorithm, we found it more useful to use the energy values to execute this task. Our observations indicate that the peak SNR (PSNR) stops increasing when the change in the energy between two consecutive iterations is below 10−1. Therefore, in practice, we suggest stopping our MG algorithm when ∥E(uk) − E(uk−1)∥2 < tol with tol = 10−1. This is roughly equivalent to stopping the algorithm when RR < 10−4. To help the reader understand our motives we show in Table 3 the data obtained from solving the benchmark problem with our MG algorithm using the following parameters : α = 1/200, β = 10−2, γ = 100, ν1 = ν2 = 10, ζ = 2, SNR = 3.5, and size = 2562.
Stopping criteria. We already have shown the convergence properties of our MG algorithm, and now we clarify the stopping criteria. Although R and RR defined above can be used to stop the algorithm, we found it more useful to use the energy values to execute this task. Our observations indicate that the peak SNR (PSNR) stops increasing when the change in the energy between two consecutive iterations is below 10−1. Therefore, in practice, we suggest stopping our MG algorithm when ∥E(uk) − E(uk−1)∥2 < tol with tol = 10−1. This is roughly equivalent to stopping the algorithm when RR < 10−4. To help the reader understand our motives we show in Table 3 the data obtained from solving the benchmark problem with our MG algorithm using the following parameters : α = 1/200, β = 10−2, γ = 100, ν1 = ν2 = 10, ζ = 2, SNR = 3.5, and size = 2562.
การแปล กรุณารอสักครู่..
