Prev Next

@(@\newcommand{\W}[1]{ \; #1 \; } \newcommand{\R}[1]{ {\rm #1} } \newcommand{\B}[1]{ {\bf #1} } \newcommand{\D}[2]{ \frac{\partial #1}{\partial #2} } \newcommand{\DD}[3]{ \frac{\partial^2 #1}{\partial #2 \partial #3} } \newcommand{\Dpow}[2]{ \frac{\partial^{#1}}{\partial {#2}^{#1}} } \newcommand{\dpow}[2]{ \frac{ {\rm d}^{#1}}{{\rm d}\, {#2}^{#1}} }@)@
First Order Reverse Mode

dw = f.Reverse(1, w)

We use @(@ F : B^n \rightarrow B^m @)@ to denote the AD function corresponding to f . The function @(@ W : B^n \rightarrow B @)@ is defined by @[@ W(x) = w_0 * F_0 ( x ) + \cdots + w_{m-1} * F_{m-1} (x) @]@ The result of this operation is the derivative @(@ dw = W^{(1)} (x) @)@; i.e., @[@ dw = w_0 * F_0^{(1)} ( x ) + \cdots + w_{m-1} * F_{m-1}^{(1)} (x) @]@ Note that if @(@ w @)@ is the i-th elementary vector , @(@ dw = F_i^{(1)} (x) @)@.

The object f has prototype
     const ADFun<
Before this call to Reverse, the value returned by
must be greater than or equal one (see size_order ).

The vector x in expression for dw above corresponds to the previous call to forward_zero using this ADFun object f ; i.e.,
f.Forward(0, x)
If there is no previous call with the first argument zero, the value of the independent variables during the recording of the AD sequence of operations is used for x .

The argument w has prototype
Vector &w
(see Vector below) and its size must be equal to m , the dimension of the range space for f .

The result dw has prototype
Vector dw
(see Vector below) and its value is the derivative @(@ W^{(1)} (x) @)@. The size of dw is equal to n , the dimension of the domain space for f .

The type Vector must be a SimpleVector class with elements of type Base . The routine CheckSimpleVector will generate an error message if this is not the case.

The file reverse_one.cpp contains an example and test of this operation. It returns true if it succeeds and false otherwise.
Input File: omh/reverse/reverse_one.omh