(X, W, b)
| 24 | |
| 25 | |
| 26 | def convpool(X, W, b): |
| 27 | # just assume pool size is (2,2) because we need to augment it with 1s |
| 28 | conv_out = tf.nn.conv2d(X, W, strides=[1, 1, 1, 1], padding='SAME') |
| 29 | conv_out = tf.nn.bias_add(conv_out, b) |
| 30 | pool_out = tf.nn.max_pool(conv_out, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME') |
| 31 | return tf.nn.relu(pool_out) |
| 32 | |
| 33 | |
| 34 | def init_filter(shape, poolsz): |