Loss functions

Differentiable financial loss functions for PyTorch training. They are drop-in PyTorch criterions usable with set_optimizer.

BaseLoss([rf, period, eps])	Base class for differentiable financial loss functions.
SharpeLoss([rf, period, eps])	Negative Sharpe ratio as a differentiable loss.
SortinoLoss([rf, period, eps])	Negative Sortino ratio as a differentiable loss.
DirectionalAccuracyLoss([rf, period, eps, ...])	Negative directional accuracy as a differentiable surrogate loss.
CalmarLoss([rf, period, eps])	Negative Calmar ratio as a differentiable loss.
OmegaLoss([threshold])	Negative Omega ratio as a differentiable loss.
HybridLoss(loss_a, loss_b[, alpha, learnable])	Convex combination of two losses: \(\alpha L_a + (1-\alpha) L_b\).