Module forward_mode

Forward-mode automatic differentiation

Forward-mode AD is efficient for computing derivatives when the number of input variables is small. It computes derivatives by propagating dual numbers through the computation graph.

Structs

ForwardADOptions

Options for forward-mode automatic differentiation

SecondOrderDual

Second-order dual number for computing exact second derivatives

Functions

forward_gradient

Compute gradient using forward-mode automatic differentiation

forward_gradient_multi

Multi-variable forward-mode gradient computation using MultiDual

forward_hessian_diagonal

Compute Hessian diagonal using forward-mode automatic differentiation

forward_hessian_diagonal_exact

Compute exact Hessian diagonal using second-order dual numbers

forward_jacobian

Forward-mode Jacobian computation for vector-valued functions

is_forward_mode_efficient

Check if forward mode is preferred for the given problem dimensions