Namespace: Models

A layer that calculates the loss between predictions and targets using a difference metric.

Nested types and modules


Difference metrics.

Functions and values

Function or valueDescription
loss lm pred target
Signature: lm:Measures -> pred:ExprT -> target:ExprT -> ExprT

Returns an expression for the loss given the loss measure lm, the predictions pred and the target values target. If the multi-class cross entropy loss measure is used then pred.[smpl, cls] must be the predicted probability that the sample belong to class cls and target.[smpl, cls] must be 1 if the sample actually belongs to class cls and 0 otherwise.

Fork me on GitHub