Hyper-Parameter | Hyper-Parameter Value |
---|---|
Number of Filters per Convolutional Layer | 350 |
Width of Convolutional Filters | 7 |
Stride of Convolutional Filters | 1 |
Number of Convolutional Layers | 5 |
Dropout for Each Convolutional Layer | 0.2 |
L2 Regularization for Each Convolutional Layer | 0.00001 |
Max-Pooling Width | 26 |
Max-Pooling Stride | 26 |
Number of Units in First Fully Connected Layer | 300 |
Optimizer | Stochastic Gradient Descent |
Learning Rate | 0.001 |
Momentum | 0.99 (Nesterov) |
Batch Size | 100 |
Class-Weighting | Fraction of Examples in the Other Class |