EntropicLoss

class pfhedge.nn.EntropicLoss(a=1.0)[source]

Creates a criterion that measures the expected exponential utility.

The loss of the profit-loss \(\text{PL}\) is given by:

\[\text{loss}(\text{PL}) = -\mathbf{E}[u(\text{PL})] \,, \quad u(x) = -\exp(-a x) \,.\]

See also

Parameters

a (float > 0, default=1.0) – Risk aversion coefficient of the exponential utility.

Shape:
  • input: \((N, *)\) where

    \(*\) means any number of additional dimensions.

  • target: \((N, *)\)

  • output: \((*)\)

Examples

>>> from pfhedge.nn import EntropicLoss
...
>>> loss = EntropicLoss()
>>> input = -torch.arange(4.0)
>>> loss(input)
tensor(7.7982)
>>> loss.cash(input)
tensor(-2.0539)