Python tf.nn.log_poisson_loss用法及代码示例

在给定 log_input 的情况下计算对数泊松损失。

用法

tf.nn.log_poisson_loss(
    targets, log_input, compute_full_loss=False, name=None
)

在假设目标具有泊松分布的情况下，给出预测和目标之间的log-likelihood 损失。警告：默认情况下，这不是确切的损失，而是损失减去常数项 [log(z!)]。这对优化没有影响，但在相对损失比较中效果不佳。要计算对数阶乘项的近似值，请指定 compute_full_loss=True 以启用斯特林近似。

为简洁起见，让 c = log(x) = log_input , z = targets 。对数泊松损失为

-log(exp(-x) * (x^z) / z!)
= -log(exp(-x) * (x^z)) + log(z!)
~ -log(exp(-x)) - log(x^z) [+ z * log(z) - z + 0.5 * log(2 * pi * z)]
    [ Note the second term is the Stirling's Approximation for log(z!).
      It is invariant to x and does not affect optimization, though
      important for correct relative loss comparisons. It is only
      computed when compute_full_loss == True. ]
= x - z * log(x) [+ z * log(z) - z + 0.5 * log(2 * pi * z)]
= exp(c) - z * c [+ z * log(z) - z + 0.5 * log(2 * pi * z)]