7. One Layer of a Convolution Network
2023. 9. 15. 13:53ㆍGoogle ML Bootcamp/4. Convolutional Neural Networks
기존 layer의 forward propagation을 생각해보면 A[l] = g(Z[l]), Z[l] = np.dot(W[l], A[l-1]) + b[l]
- W[l]이 filter가 된다. 이젠 np.dot 처럼 행렬곱이 convolution 연산을 수행할 뿐.
- Z[l] = W * A[l-1] + b , 이때 *는 convolution operation.
'Google ML Bootcamp > 4. Convolutional Neural Networks' 카테고리의 다른 글
| 9. Pooling Layers (0) | 2023.09.15 |
|---|---|
| 8. Simple Convolutional Network Example (0) | 2023.09.15 |
| 6. Convolutions Over Volume (0) | 2023.09.15 |
| 5. Strided Convolutions (0) | 2023.09.15 |
| 4. Padding (0) | 2023.09.15 |