Forward Mode Automatic Differentiation

Concept

A mode of automatic differentiation where gradients are computed forward through the computation graph, which is inefficient for machine learning with many parameters.

Mentioned in 1 video