16.01.2015 Views

GAMS — The Solver Manuals - Available Software

GAMS — The Solver Manuals - Available Software

GAMS — The Solver Manuals - Available Software

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

334 MINOS<br />

1 Introduction<br />

This document describes the <strong>GAMS</strong> interface to MINOS which is a general purpose nonlinear programming solver.<br />

<strong>GAMS</strong>/MINOS is a specially adapted version of the solver that is used for solving linear and nonlinear programming<br />

problems in a <strong>GAMS</strong> environment.<br />

<strong>GAMS</strong>/MINOS is designed to find solutions that are locally optimal. <strong>The</strong> nonlinear functions in a problem must<br />

be smooth (i.e., their first derivatives must exist). <strong>The</strong> functions need not be separable. Integer restrictions cannot<br />

be imposed directly.<br />

A certain region is defined by the linear constraints in a problem and by the bounds on the variables. If the<br />

nonlinear objective and constraint functions are convex within this region, any optimal solution obtained will be a<br />

global optimum. Otherwise there may be several local optima, and some of these may not be global. In such cases<br />

the chances of finding a global optimum are usually increased by choosing a staring point that is sufficiently close,<br />

but there is no general procedure for determining what close means, or for verifying that a given local optimum<br />

is indeed global.<br />

<strong>GAMS</strong> allows you to specify values for many parameters that control <strong>GAMS</strong>/MINOS, and with careful experimentation<br />

you may be able to influence the solution process in a helpful way. All MINOS options available through<br />

<strong>GAMS</strong>/MINOS are summarized at the end of this document.<br />

2 How to Run a Model with <strong>GAMS</strong>/MINOS<br />

MINOS is capable of solving models of the following types: LP, NLP, DNLP and RMINLP. If MINOS is not<br />

specified as the default LP, NLP, DNLP or RMINLP solver, then the following statement can be used in your<br />

<strong>GAMS</strong> model:<br />

or<br />

option nlp=minos; { or lp or dnlp or rminlp }<br />

option nlp=minos55; { or lp or dnlp or rminlp }<br />

It should appear before the solve statement.<br />

This will invoke MINOS 5.5. In some cases an older version of MINOS, version 5.4 is more efficient than the<br />

newer version. MINOS 5.4 can be selected by:<br />

option nlp=minos5; { or lp or dnlp or rminlp }<br />

To be complete, we mention that this can be also specified on the command line, as in:<br />

> gams camcge nlp=minos<br />

This will override the global default, but if an algorithm option has been specified inside the model, then that<br />

specification takes precedence.<br />

3 Overview of <strong>GAMS</strong>/MINOS<br />

<strong>GAMS</strong>/MINOS is a system designed to solve large-scale optimization problems expressed in the following form:<br />

NLP<br />

minimize F (x) + c T x + d T y (1)<br />

x,y<br />

subject to f(x) + A 1 y ∼ b 1 (2)<br />

A 2 x ( + A 3 y ∼ b 2 (3)<br />

x<br />

l ≤ ≤ u (4)<br />

y)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!