tf.keras.backend.batch_dot函数

tf.keras.backend.batch_dot(
    x,
    y,
    axes=None
)

定义在：tensorflow/python/keras/backend.py。

批量化的点积。

当x和y是批量数据时，batch_dot用于计算x和y的点积，即shape为（batch_size，:)。 batch_dot产生维度小于输入的张量或变量。如果维度数减少到1，我们使用expand_dims确保ndim至少为2。

参数：

• x：ndim >= 2的Keras张量或变量。
• y：ndim >= 2的Keras张量或变量。
• axes：具有目标维度的（或单个）int列表。axes[0]和axes[1]的长度应该是相同的。

返回：

A tensor with shape equal to the concatenation of `x`'s shape
(less the dimension that was summed over) and `y`'s shape
(less the batch dimension and the dimension that was summed over).
If the final rank is 1, we reshape it to `(batch_size, 1)`.

例子：假设x = [[1, 2], [3, 4]]，y = [[5, 6], [7, 8]] ，其中batch_dot(x, y, axes=1) = [[17, 53]]是x.dot(y.T)的主对角线，虽然我们没有必要计算非对角线元素。

Shape inference:
Let `x`'s shape be `(100, 20)` and `y`'s shape be `(100, 30, 20)`.
If `axes` is (1, 2), to find the output shape of resultant tensor,
    loop through each dimension in `x`'s shape and `y`'s shape:

* `x.shape[0]` : 100 : append to output shape
* `x.shape[1]` : 20 : do not append to output shape,
    dimension 1 of `x` has been summed over. (`dot_axes[0]` = 1)
* `y.shape[0]` : 100 : do not append to output shape,
    always ignore first dimension of `y`
* `y.shape[1]` : 30 : append to output shape
* `y.shape[2]` : 20 : do not append to output shape,
    dimension 2 of `y` has been summed over. (`dot_axes[1]` = 2)
`output_shape` = `(100, 30)`

    >>> x_batch = K.ones(shape=(32, 20, 1))
    >>> y_batch = K.ones(shape=(32, 30, 20))
    >>> xy_batch_dot = K.batch_dot(x_batch, y_batch, axes=[1, 2])
    >>> K.int_shape(xy_batch_dot)
    (32, 1, 30)