Reverse Mode Hessian Sparsity Patterns

@(@\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}} }@)@This is cppad-20221105 documentation. Here is a link to its current documentation . Reverse Mode Hessian Sparsity Patterns
Syntax

f.rev_hes_sparsity(

    select_range, transpose, internal_bool, pattern_out

)

Purpose
We use @(@ F : \B{R}^n \rightarrow \B{R}^m @)@ to denote the AD function corresponding to the operation sequence stored in f . Fix @(@ R \in \B{R}^{n \times \ell} @)@, @(@ s \in \B{R}^m @)@ and define the function @[@ H(x) = ( s^\R{T} F )^{(2)} ( x ) R @]@ Given a sparsity pattern for @(@ R @)@ and for the vector @(@ s @)@, rev_hes_sparsity computes a sparsity pattern for @(@ H(x) @)@.

x
Note that the sparsity pattern @(@ H(x) @)@ corresponds to the operation sequence stored in f and does not depend on the argument x .

BoolVector
The type BoolVector is a SimpleVector class with elements of type bool.

SizeVector
The type SizeVector is a SimpleVector class with elements of type size_t.

f
The object f has prototype



    ADFun<Base> f

R
The sparsity pattern for the matrix @(@ R @)@ is specified by pattern_in in the previous call



    f.for_jac_sparsity(

        pattern_in, transpose, dependency, internal_bool, pattern_out

)

select_range
The argument select_range has prototype



    const BoolVector& select_range

It has size @(@ m @)@ and specifies which components of the vector @(@ s @)@ are non-zero; i.e., select_range[i] is true if and only if @(@ s_i @)@ is possibly non-zero.

transpose
This argument has prototype



    bool transpose

See pattern_out below.

internal_bool
If this is true, calculations are done with sets represented by a vector of boolean values. Otherwise, a vector of sets of integers is used. This must be the same as in the previous call to f.for_jac_sparsity .

pattern_out
This argument has prototype



    sparse_rc<SizeVector>& pattern_out

This input value of pattern_out does not matter. If transpose it is false (true), upon return pattern_out is a sparsity pattern for @(@ H(x) @)@ (@(@ H(x)^\R{T} @)@).

Sparsity for Entire Hessian
Suppose that @(@ R @)@ is the @(@ n \times n @)@ identity matrix. In this case, pattern_out is a sparsity pattern for @(@ (s^\R{T} F) F^{(2)} ( x ) @)@.

Example
The file rev_hes_sparsity.cpp contains an example and test of this operation.

Input File: include/cppad/core/rev_hes_sparsity.hpp