tf.reduce_logsumexp 函数介绍

tf.reduce_logsumexp 函数

reduce_logsumexp(
    input_tensor,
    axis=None,
    keep_dims=False,
    name=None,
    reduction_indices=None
)

沿着axis指定的轴计算。如果axis=None，那么将计算整个input_tensor张量的值，并返回一个标量。

例如

x = tf.constant([[0., 0., 0.], [0., 0., 0.]])
tf.reduce_logsumexp(x)  # log(6)
tf.reduce_logsumexp(x, 0)  # [log(2), log(2), log(2)]
tf.reduce_logsumexp(x, 1)  # [log(3), log(3)]
tf.reduce_logsumexp(x, 1, keep_dims=True)  # [[log(3)], [log(3)]]
tf.reduce_logsumexp(x, [0, 1])  # log(6)

x = tf.constant([[0., 1., 2.], [0., 1., 2.]])
y1 = tf.reduce_logsumexp(x)             # 3.1007533
y2 = tf.reduce_logsumexp(x, 0)          # array([log(2), log(2e), log(4e)], dtype=float32)
y3 = tf.reduce_logsumexp(x, 1)          # array([log(e^2+e+1), log(e^2+e+1)], dtype=float32)
y4 = tf.reduce_logsumexp(x, [0, 1])     # log(2*e^2+2*e+2)
y5 = tf.reduce_logsumexp(x)             # log(2*e^2+2*e+2)