Function focal_loss 

pub fn focal_loss(
    graph: &mut Graph,
    prediction: NodeId,
    target: NodeId,
    alpha: f32,
    gamma: f32,
) -> Result<NodeId, ModelError>

Focal loss for imbalanced classification.

FL = -alpha * (1 - p_t)^gamma * log(p_t) averaged over elements. prediction should be sigmoid probabilities, target binary labels.