Skip to main content

Table 2 The effect of optimizer in unimodal classifier of music or video network

From: Deep learning-based late fusion of multimodal information for emotion classification of music video

Optimizer 1D Music CNN 2D Music CNN C3D Video Network I3D Video Network
Adam LR: 0.001
Test Accuracy: 0.4453
F1-score: 0.41
ROC AUC Score: 0.757
LR: 0.001
Test Accuracy: 0.6539
F1-score: 0.65
ROC AUC Score: 0.916
LR: 0.00001
Test Accuracy: 0.6423
F1-score: 0.64
ROC AUC Score: 0.898
LR: 0.0001
Test Accuracy: 0.6622
F1-score: 0.66
ROC AUC Score: 0.899
SGD LR: 0.001 M: 0.5
Test Accuracy: 0.4238
F1-score: 0.37
ROC AUC Score: 0.734
LR: 0.001 M: 0.5
Test Accuracy: 0.7251
F1-score: 0.72
ROC AUC Score: 0.934
LR: 0.00001 M: 0.0
Test Accuracy: 0.6440
F1-score: 0.64
ROC AUC Score: 0.890
LR: 0.001 M: 0.5
Test Accuracy: 0.6026
F1-score: 0.59
ROC AUC Score: 0.879
RMSprop LR: 0.001
Test Accuracy: 0.4304
F1-score: 0.39
ROC AUC Score: 0.748
LR: 0.001
Test Accuracy: 0.6887
F1-score: 0.68
ROC AUC Score: 0.921
LR: 0.001
Test Accuracy: 0.5678
F1-score: 0.56
ROC AUC Score: 0.848
LR: 0.001
Test Accuracy: 0.5281
F1-score: 0.53
ROC AUC Score: 0.832
  1. The bold number represent the highest evaluation score
  2. LR, learning rate and M, momentum