CNN Model Hyperparameter Tuning Algorithms:
- Bayesian Optimization integrated with Differential Evolution, Harmony Search and Particle Swarm Optimization
- Differential Evolution
- Harmony Search
- Particle Swarm Optimization
Dataset:
Click to expand/collapse the table
| Aged DE Tuned | Aged HS Tuned | Aged PSO Tuned | Aged Byopt DE Tuned | Aged Byopt HS Tuned | Aged Byopt PSO Tuned | DE Tuned | HS Tuned | PSO Tuned | Byopt DE Tuned | Byopt HS Tuned | Byopt PSO Tuned | Naive | Voting | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Layer1_filter | 32.00 | 32.00 | 64.00 | 32.00 | 64.00 | 64.00 | 64.00 | 32.00 | 64.00 | 32.00 | 32.00 | 64.00 | 32.00 | 32 |
| Layer1_act | relu | relu | relu | sigmoid | sigmoid | relu | relu | relu | relu | sigmoid | relu | relu | relu | relu |
| Layer2_filter | 32.00 | 32.00 | 32.00 | 32.00 | 32.00 | 64.00 | 32.00 | 32.00 | 64.00 | 32.00 | 64.00 | 64.00 | 32.00 | 32 |
| Layer2_act | sigmoid | relu | sigmoid | sigmoid | relu | sigmoid | sigmoid | relu | sigmoid | relu | sigmoid | sigmoid | relu | sigmoid |
| Drop1 | 0.45 | 0.40 | 0.25 | 0.50 | 0.50 | 0.50 | 0.40 | 0.25 | 0.25 | 0.25 | 0.25 | 0.30 | 0.25 | 0.25 |
| Layer3_filter | 32.00 | 128.00 | 32.00 | 32.00 | 128.00 | 128.00 | 64.00 | 64.00 | 64.00 | 128.00 | 128.00 | 32.00 | 64.00 | 128 |
| Layer3_act | sigmoid | relu | relu | relu | sigmoid | sigmoid | relu | relu | sigmoid | relu | sigmoid | relu | relu | relu |
| Layer4_filter | 32.00 | 64.00 | 64.00 | 32.00 | 32.00 | 32.00 | 32.00 | 64.00 | 64.00 | 32.00 | 32.00 | 32.00 | 64.00 | 32 |
| Layer4_act | relu | relu | relu | sigmoid | sigmoid | relu | sigmoid | relu | relu | relu | sigmoid | relu | relu | relu |
| Drop2 | 0.45 | 0.50 | 0.50 | 0.25 | 0.50 | 0.35 | 0.50 | 0.25 | 0.30 | 0.25 | 0.50 | 0.40 | 0.25 | 0.5 |
| Layer5_filter | 256.00 | 64.00 | 256.00 | 256.00 | 256.00 | 64.00 | 256.00 | 256.00 | 128.00 | 256.00 | 128.00 | 64.00 | 128.00 | 256 |
| Layer5_act | relu | sigmoid | sigmoid | sigmoid | relu | sigmoid | relu | relu | sigmoid | sigmoid | sigmoid | sigmoid | relu | sigmoid |
| Layer6_filter | 64.00 | 64.00 | 256.00 | 64.00 | 64.00 | 256.00 | 64.00 | 64.00 | 256.00 | 256.00 | 64.00 | 256.00 | 128.00 | 64 |
| Layer6_act | relu | sigmoid | relu | relu | relu | relu | relu | relu | sigmoid | sigmoid | sigmoid | relu | relu | relu |
| Drop3 | 0.50 | 0.50 | 0.50 | 0.50 | 0.50 | 0.25 | 0.35 | 0.50 | 0.50 | 0.35 | 0.45 | 0.25 | 0.25 | 0.5 |
| Layer7_units | 256.00 | 1,024.00 | 512.00 | 1,024.00 | 256.00 | 1,024.00 | 128.00 | 512.00 | 1,024.00 | 1,024.00 | 512.00 | 512.00 | 128.00 | 1024 |
| Layer7_act | relu | sigmoid | sigmoid | sigmoid | relu | relu | sigmoid | relu | relu | relu | sigmoid | relu | relu | relu |
| Drop4 | 0.25 | 0.25 | 0.25 | 0.25 | 0.25 | 0.25 | 0.45 | 0.40 | 0.30 | 0.50 | 0.25 | 0.40 | 0.25 | 0.25 |
| optimizer | nadam | nadam | sgd | adam | adam | sgd | rmsprop | nadam | nadam | sgd | sgd | nadam | adam | nadam |
| epochs | 30.00 | 16.00 | 30.00 | 30.00 | 23.00 | 15.00 | 24.00 | 30.00 | 24.00 | 30.00 | 26.00 | 22.00 | 30.00 | 30 |
| batch_size | 8.00 | 16.00 | 8.00 | 8.00 | 16.00 | 8.00 | 32.00 | 16.00 | 32.00 | 8.00 | 8.00 | 64.00 | 16.00 | 8 |
| learning_rate | 0.001217 | 0.001000 | 0.010000 | 0.001000 | 0.001084 | 0.010000 | 0.001018 | 0.001147 | 0.001000 | 0.010000 | 0.010000 | 0.001508 | 0.001 | 0.001 |
| best_fitness | 0.94 | 0.87 | 0.90 | 0.86 | 0.86 | 0.88 | 0.88 | 0.89 | 0.89 | 0.90 | 0.85 | 0.85 |
Click to expand/collapse the table
| Model | Best Accuracy | Max Accuracy | Min Accuracy | Mean Accuracy | PI(Best/N) | PI(Mean/N) | PI(Max/N) | PRD(Max/Best) | Skewness | Kurtosis | Std |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Aged DE Tuned | 94% | 88% | 62% | 79.706666% | 10.588235% | 7.998735% | 3.529408% | 6.590000% | -0.563683 | 1.050164 | 0.037693 |
| Aged HS Tuned | 87% | 86% | 51% | 77.773333% | 2.352941% | 5.379161% | 1.176469% | 1.160000% | -1.318221 | 3.678665 | 0.048656 |
| Aged PSO Tuned | 90% | 91% | 77% | 83.883333% | 5.882353% | 13.657919% | 7.058824% | 1.100000% | -0.174784 | -0.238688 | 0.025726 |
| Aged Byopt DE Tuned | 86% | 86% | 65% | 78.630000% | 1.176471% | 6.539903% | 1.176469% | 0.000000% | -0.654741 | 1.080759 | 0.035282 |
| Aged Byopt HS Tuned | 86% | 85% | 57% | 77.023333% | 1.176471% | 4.362946% | 0.000000% | 1.170000% | -1.385112 | 3.628307 | 0.044274 |
| Aged Byopt PSO Tuned | 88% | 91% | 71% | 80.616666% | 3.529412% | 9.231742% | 7.058824% | 3.350000% | -0.183008 | 0.020675 | 0.034462 |
| DE Tuned | 88% | 88% | 55% | 78.386666% | 3.529412% | 6.210198% | 3.529408% | 0.000000% | -1.246771 | 2.593043 | 0.048802 |
| HS Tuned | 89% | 88% | 49% | 78.243333% | 4.705882% | 6.015988% | 3.529408% | 1.130000% | -1.900475 | 9.056414 | 0.047854 |
| PSO Tuned | 89% | 87% | 43% | 75.363333% | 4.705882% | 2.113725% | 2.352939% | 2.270000% | -1.210384 | 3.374937 | 0.059881 |
| Byopt DE Tuned | 90% | 90% | 74% | 81.613333% | 5.882353% | 10.582177% | 5.882347% | 0.000000% | -0.197353 | -0.005680 | 0.027778 |
| Byopt HS Tuned | 85% | 85% | 64% | 77.053333% | 0.000000% | 4.403595% | 0.000000% | 0.000000% | -0.413858 | 0.649220 | 0.033015 |
| Byopt PSO Tuned | 85% | 90% | 31% | 75.266667% | 0.000000% | 1.982747% | 5.882347% | 5.710000% | -1.670602 | 4.609862 | 0.078290 |
| Voting | 91% | 91% | 56% | 80.013333% | 7.058827% | 8.414254% | 7.058824% | 0.000000% | -0.879668 | 2.813564 | 0.043524 |
| Naive | 85% | 85% | 60% | 73.803333% | 0.000000% | 0.000000% | 0.000000% | 0.000000% | -0.193603 | 0.310538 | 0.042613 |