tf.keras.layers运算

tf.keras.layers.Permute

tf.keras.layers.Permute(
    dims, **kwargs
)

使用案例：

model = Sequential()
model.add(Permute((2, 1), input_shape=(10, 64)))
# now: model.output_shape == (None, 64, 10)
# note: `None` is the batch dimension

参数：

dims : Tuple of integers. Permutation pattern does not include the samples dimension. Indexing starts at 1. For instance, (2, 1) permutes the first and second dimensions of the input.

tf.keras.layers.Multiply

函数原型：

tf.keras.layers.Multiply(
    **kwargs
)

使用案例：

tf.keras.layers.Multiply()([np.arange(5).reshape(5, 1),
                            np.arange(5, 10).reshape(5, 1)])

输出：

tf.Tensor(
[[ 0]
 [ 6]
 [14]
 [24]
 [36]], shape=(5, 1), dtype=int32)

由此可见，tf.keras.layers.Multiply就是相同shape的tensor对应位置元素相乘。

tf.keras.layers.Reshape

函数原型

tf.keras.layers.Reshape(
    target_shape, **kwargs
)

使用案例：

n = np.arange(32).reshape(2, 16)
k = tf.convert_to_tensor(n)
xx = tf.keras.layers.Reshape((8, 2))(k)
xn = tf.keras.layers.Reshape((4, 4))(xx)

n1 = np.arange(32).reshape(16, 2)
k1 = tf.convert_to_tensor(n1)
xx1 = tf.keras.layers.Reshape((1, 2))(k1)

可以看出，输出为(batch_size,) + target_shape
第一维肯定是batch_size，第二维开始按行的顺序重新排列。重新排列时，按照维度从前到后的顺序重新排列。每次reshape都按照这个顺序，乱不了。

model = tf.keras.Sequential()
model.add(tf.keras.layers.Reshape((3, 4), input_shape=(12,)))
# model.output_shape == (None, 3, 4), `None` is the batch size.
model.output_shape

输出：（None, 3, 4）

tf.keras.layers.RepeatVector

函数原型：

tf.keras.layers.RepeatVector(
    n, **kwargs
)

使用案例：

model = Sequential()
model.add(Dense(32, input_dim=32))
# now: model.output_shape == (None, 32)
# note: `None` is the batch dimension

model.add(RepeatVector(3))
# now: model.output_shape == (None, 3, 32)

Input shape:
2D tensor of shape (num_samples, features).

Output shape:
3D tensor of shape (num_samples, n, features).