Fig. 7
From: Partitioning multi-layer edge network for neural network collaborative computing
Comparison of convergence speed between OEGA and other algorithms. OEGA converges faster than the other algorithms and has converged to the current optimal solution after about 200 ~ 400 iterations, while SGA and EGA need about 400 ~ 1000 iterations to converge, and the convergence value of OEGA is better under the statistical final delay convergence value of the three neural networks, showing that compared with SGA and EGA, OEGA has faster convergence speed and better convergence under the same number of iterations