Table 3 Applied parameters for training the proposed models.
From: Speech emotion recognition with light weight deep neural ensemble model using hand crafted features
Parameter | Value |
|---|---|
Batch size | 64 |
Epochs | 100 |
Objective function | Cross-Entropy (Categorical) |
Final layer activation | Softmax function |
Adam optimizer | 0.001 |
LR adjustment | monitor=’validation_accuracy’, patience=3, reduction_factor=0.5, minimum_LR=0.00001 |
Early stopping | monitor=’val_accuracy’, patience=5 |
Kernel size | ”5x1” for first 3 Convolutional Layers ”3x1” for the rest |
Pool size | ”2x2” for first 4 MaxPooling Layers ”2x1” for the last |
Activation | Relu (Rectified Linear Unit) |
Padding | Same |
Dropout rate | 20% |
