source: coopr.dist/trunk/doc/coopr/pyomo-advanced.tex @ 1808

Revision 1808, 1.7 KB checked in by wehart, 4 years ago (diff)

Reorganization of documentation directories.

1\chapter{Advanced Pyomo Examples}
3\section{XXX Advanced Pyomo Examples}
5\subsection{Parallel Benders Decomposition}
7TODO: Does the following paragraph go here?  I think that Dave's talking
8about the abilities that we're leveraging in PH, but we haven't discussed
9that in this paper.  Perhaps this should go back into the introduction,
10but if so then the intent needs to be clarified for me (BILL).
12An important consequence of the design using Python and integration
13with the Coopr environment is that modularity is fully supported over
14a range of abstraction.  At one extreme, the model elements can be
15manipulated explicitly by specifying their names and the values of
16their indexes. This sort of reference can be made more abstract, as
17is the case with algebraic modeling languages, by specifying various
18types of named sets so that the dimensions and details of the data
19can be separated from the specification of the model. Separation of an
20abstract declarative model from the data specification is a hallmark of
21structure modeling techniques for efficient modeling \citep{Geoffrion}.
22At the other extreme, elements of a mathematical program can be treated in
23their fully canonical form as is supported by callable solver libraries.
24Methods can be written that operate, for example, on objective functions
25or constraints in a fully general way. This capability is a fundamental
26tool for general algorithm development and extension \citep{Marsten}.
27Pyomo provides the full continuum of abstraction between these two
28extremes to support modeling and development.  Furthermore, methods are
29extensible via overloading of all defined operations. Both modelers and
30developers can alter the behavior of the package or add new functionality.
Note: See TracBrowser for help on using the repository browser.