Module reverse 

Reverse mode automatic differentiation (backpropagation)

Reverse mode AD is efficient for computing gradients when the number of outputs is small compared to the number of inputs.

Structs

ReverseAD

Reverse mode automatic differentiation engine

Tape

Reverse mode AD tape for recording operations

TapeNode

Node in the computation graph

Enums

CheckpointStrategy

Checkpointing strategy for memory-efficient gradient computation

Operation

Operations that can be recorded on the tape

Functions

reverse_gradient

Compute gradient using reverse mode AD (convenience function)

reverse_jacobian

Compute Jacobian using reverse mode AD (convenience function)