Research Article
[Retracted] Gradient Descent Optimization in Deep Learning Model Training Based on Multistage and Method Combination Strategy
Table 13
Performance of the proposed method, SGD with Momentum and decay.
| | ResNet-20 on Cafri-10 | LSTM on IMDB | Val-loss | Val-acc | Val-loss | Val-acc |
| (SGD + M + d) + SGD | 0.6324 | 0.8340 | 0.3736 | 0.8366 | (SGD + M + d) + (SGD + M) | 0.8533 | 0.7737 | 0.3770 | 0.8369 | (SGD + M + d) + (SGD + d) | 0.6485 | 0.8278 | 0.3970 | 0.8234 | (SGD + M + d) + (SGD + M + d) | 0.8289 | 0.7868 | 0.4274 | 0.8036 | (SGD + M + d) + RMSprop | 0.9795 | 0.7593 | 0.4949 | 0.8245 | (SGD + M + d) + (RMSprop + d) | 0.8371 | 0.7884 | 0.4587 | 0.8279 | (SGD + M + d) + Adam | 1.2722 | 0.7058 | 0.8569 | 0.8091 | (SGD + M + d) + (Adam + d) | 0.7481 | 0.8112 | 0.9056 | 0.8070 |
|
|
The bold values represent the best results.
|