gaw.n

较好的一种新的关于遗传算法的源程序

.fo ''%''
.tp
.m4 0.7i
.po +0.75i
.nr pp 11
.nr tp 12
.nr sp 12
.sz +2
.ce
.uh "Genetic Algorithm Workbench Documentation"
.sz -2







.lp
\fIThe Genetic Algorithm Workbench program and its documentation are
copyright and may not be copied or distributed without written permission
from the author with the following exceptions:

.ip \fI(1)
\fICopies of the software and this documentation may be made and passed on to
any third party provided that all the files on the distribution disk are
distributed together in unmodified form, and providing that no profit is
made from such distribution.

.ip \fI(2)
\fIA reasonable number of copies may be made of the software for the purpose
of archiving to guard against corruption of the working copy of the software.

.lp
\fIThe software can be used without restriction or payment, but you are
encouraged to send an appropriate contribution in sterling to the author if
you feel that the program has been of use.
See section 4 for the author's address.

.lp
\fINo warranty is given that this software is fit for any purpose, nor that it
will perform as described in this manual.
You use it entirely at your own risk.\fR




.(l
\fICopyright (C) Mark Hughes 1989. All rights reserved.

Last Change:  30 November 1989.\fR
.)l
.bp
.uh "CONTENTS"

.(l
1.	Introduction

2.	Purpose

3.	Using the Genetic Algorithm Workbench Program

	3.1	Hardware Requirements
	3.2	Running the Program
	3.3	Screen Display
	3.4	Menu Commands
	3.5	Program Control Variables

4.	Bibliography

5.	Appendix A -   Workbench Algorithms

	5.1	Solving Problems with a Genetic Algorithm
	5.2	Genetic Coding
	5.3	Genetic Algorithm Implementation
	5.4	Summary of Algorithm Input Variables

		5.4.1	Population
		5.4.2	Fitness Scaling
		5.4.3	Breeder Selection
		5.4.4	Generation Gap
		5.4.5	Mates Selection
		5.4.6	Mating
		5.4.7	Crossover
		5.4.8	Mutation Probability
		5.4.9	Dispersal
		5.4.10	Crowding Factor
		5.4.11	Elitism
		5.4.12	Sacrifice Selection

	5.5	Output Variables
	5.6	References

6.	Appendix B -   A Typical Session Using the Workbench

7.	Appendix C -   Problems and How to Fix Them

8.	Appendix D -   General Introduction to Genetic Algorithms

9.	Appendix E -   Main Command Menu
.bp
.lp
.sh 1 "Introduction"

.lp
This is a user's manual for the Genetic Algorithm Workbench program
which is an interactive tool for demonstrating and experimenting with genetic
algorithms using an IBM compatible personal computer.
This is not a set of subroutines for inclusion in your own programs.
If that is what you require, see the bibliography for details of GENESIS.

.lp
The manual commences with a description of the purpose of the Workbench in
section 2.

.lp
Section 3 then describes how to use the program and gives details of
hardware requirements, instructions for running the program, an explanation
of the screen display, and explains how to control the program using the
command menu and program control variables.

.lp
Section 4 contains a short bibliography.

.lp
Appendix A describes the detailed operation of the genetic algorithm employed
by the Workbench and the effect of each algorithm input variable.
It gives details of where to find further information about the theory of
the different aspects of the genetic algorithm which are described.
An explanation of each output variable displayed on the screen is also given.

.lp
Appendix B contains a short step by step example of using the Workbench.

.lp
Appendix C may help if you encounter problems trying to use the Workbench.

.lp
Appendix D includes an article as a general introduction to genetic
algorithms and their applications.

.lp
Appendix E is a brief summary of the Workbench command menu.

.sh 1 "Purpose"

.lp
The purpose of the Workbench program is to allow experimentation with
search and optimisation algorithms.
It is primarily a tool for experimenting with different types of genetic
algorithm, but is also intended for use in comparing genetic algorithms with
other techniques although so far only the genetic algorithm has been
implemented.

.lp
A genetic algorithm is a method for finding the peaks of difficult functions,
and the Workbench program is both for demonstrating genetic algorithms and
for evaluating their performance.

.lp
The idea is that you provide a function, the "Target Function" and see how
quickly the selected algorithm is able to find the peak value, or indeed if
it succeeds at all.
You can vary the details of the algorithm used by tweaking several
numeric control parameters and selecting different types of operator
employed by the algorithm.

.sh 1 "Using the Genetic Algorithm Workbench Program"

.lp
The following sections describe hardware required and provide instructions 
for starting the Workbench program.
The screen display is then explained followed by details of how to control
the program, and finally a description of menu commands is given.

.lp
Appendix B describes a typical session using the Workbench, and will also be
useful while learning how to use the program.

.sh 2 "Hardware Requirements"

.lp
To run the genetic algorithm Workbench, you require

.(l
IBM PC/XT/AT or compatible
EGA display
Microsoft compatible mouse
.)l

.lp
The Workbench has been tested on MS-DOS version 3.3 but should work on
most versions of MS-DOS and PC-DOS.
Reports of problems with other versions will be appreciated.

.lp
The program will work on a VGA mode screen, but unless you can put it in EGA
emulation mode, the program will only use the lower two thirds of the
screen, giving a squashed display.

.sh 2 "Running the Program"

.lp
Check that your system hardware is suitable for running the genetic
algorithm workbench (see above).

.lp
Switch on the computer and wait for your MS-DOS prompt to appear on the screen.
If you have a display adapter board that can emulate several display
modes, ensure that it is in EGA mode.
This might be done by setting configuration switches on the adapter
card prior to switching your computer on, or by means of a special
command provided with your display adapter.
Refer to your display adapter manual for details of how to set it to EGA mode.

.lp
Now, ensure that your mouse driver is loaded.
On some systems a command such as "MOUSE" is provided to load the driver, on
others a command must be inserted into your config.sys file.
Refer to the documentation for your mouse on how to install the mouse driver.

.lp
Now the final test.
Insert the Genetic Algorithm Workbench disk into drive a: and type

.(l
a:gaw
.)l

.lp
and press the <ENTER> key.

.lp
The program takes a little while to load but you should see a display
similar to that shown in figure 1 (see next section), and if you move your
mouse, the mouse cursor should be visible and will change as you move it
over different areas of the screen.

.lp
If you don't think everything is as described here, refer to appendix C
which describes possible problems and how to deal with them.

.sh 2 "Screen Display"

.lp
Figure 1 shows a screen display similar to the one you should see when
the Workbench is first started.
The main features of the display are as follows.

.ip "\fICommand Menu\fR"
This is a menu of commands shown at the top left of the screen which are
activated using the mouse.
Each command is highlighted when the mouse cursor overlays it, and is executed
if the left mouse button is pressed while the command is highlighted.
See section 3.5, Program Control for details of these commands.

.ip "\fIAlgorithm Control Chapter\fR"
This is the large box which spans the top of the screen to the right of the
command menu.
It is called a \fIchapter\fR because it can contain several \fIpages,\fR
only one of which is visible at one time.
Initially, it displays a page called "Simple Genetic Algorithm" which shows
a number of input variables and their values, which are used to control
the operation of this algorithm.
Pages can be flipped through, forwards or backwards, by clicking the left
mouse button on the arrows in the top right hand corner of the chapter.
Each page is described below.

.ip "\fISimple Genetic Algorithm Page\fR"
This is the box within the algorithm control chapter entitled "Simple Genetic
Algorithm".
(Note that it may be necessary to select this page using the mouse before
it is displayed within the chapter box - see "Algorithm Control Chapter" above.)
This page is normally displayed when the Workbench is first started and
lists a number of control inputs to the genetic algorithm together with
their current values.
The page shows two columns of input variables.
Each variable displayed shows its name to the left, a pair of arrows in the
middle, and the variable's current value to the right.
Note that many of the variable values are text strings.
You can alter the value of any of these variables by clicking the left mouse
button on the up or down arrows to the left of each value.
The meaning of each variable is described in appendix A.

.ip "\fIGeneral Program Control Variables Page\fR"
This is the box within the algorithm control chapter entitled "General Program
Control Variables".
(Note that it may be necessary to select this page using the mouse before
it is displayed within the chapter box - see "Algorithm Control Chapter" above.)
This page contains variables related to general program operation rather
than a specific algorithm.
The meaning of each program control variable is described in section 3.5.

.ip "\fIOutput Variables Box\fR"
This is the box at the bottom right of the screen which contains the current
values of a number of variables relating to the current algorithm.
These variables cannot be altered by the user; they indicate the current
state of the algorithm.
Each output variable is described in appendix A.

.ip "\fITarget Function Graph\fR"
This is the graph labelled "Target Function" and is used for entry and display
of the function to be solved.
Using the Workbench involves drawing functions on this graph which are then
solved using the selected algorithm.
After a function has been provided, a small red triangle on the x axis marks
the highest peak, which is the target for the algorithm to find.

.ip "\fIPopulation Distribution Histogram\fR"
This is the plot entitled "Population Distribution" immediately below the
target function graph and shows the distribution of organisms by value of
x for the genetic algorithm.

.ip "\fIOutput Graph\fR"
This is the graph labelled "Output Plot" and is used to display plots of
various output variables against time.

.ip "\fIAxis Value Box\fR"