Table 2 Architecture of the convolutional neural network used in this work
From: Insightful classification of crystal structures using deep learning
Layer type | Specifications |
|---|---|
Convolutional layer | (Kernel: 7 × 7; 32 filters) |
Convolutional layer | (Kernel: 7 × 7; 32 filters) |
Max pooling layer | (Pool size: 2 × 2; stride: 2 × 2) |
Convolutional layer | (Kernel: 7 × 7; 16 filters) |
Convolutional layer | (Kernel: 7 × 7; 16 filters) |
Max pooling layer | (Pool size: 2 × 2; stride: 2 × 2) |
Convolutional layer | (Kernel: 7 × 7; 8 filters) |
Convolutional layer | (Kernel: 7 × 7; 8 filters) |
Fully connected layer + dropout | (Size: 128; dropout: 25%) |
Batch normalization | (Size: 128) |
Softmax | (Size: 7) |