Browse Prior Art Database

Automatic Selection of Breeding Procedure in Mutation System

IP.com Disclosure Number: IPCOM000113401D
Original Publication Date: 1994-Aug-01
Included in the Prior Art Database: 2005-Mar-27
Document File: 2 page(s) / 92K

Publishing Venue

IBM

Related People

Simmons, J: AUTHOR [+2]

Abstract

Disclosed is a modification to a basic mutator system whereby certain algorithms can be used to select one of a number of breeding routines from an analysis of the users previous judgements. The method has two advantages. It makes mutation easier to learn and removes some of the process oriented decisions from the user.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Automatic Selection of Breeding Procedure in Mutation System

      Disclosed is a modification to a basic mutator system whereby
certain algorithms can be used to select one of a number of breeding
routines from an analysis of the users previous judgements.   The
method has two advantages.  It makes mutation easier to learn and
removes some of the process oriented decisions from the user.

      The problem is that Mutator technology is designed to let a
user interact with a computer by a series of subjective judgements.
There are several algorithms used within Mutator (mutation, steering,
marriage in different forms), and the user decides which one is most
appropriate at each step.   This decision distracts the user from the
application judgements (cognitive overload).

      In a mutation system for creating new gene values the following
algorithms are used:

o   Mutation - in which child gene values are generated by random
    mutation of parent gene values.

o   Steering - in which child gene values are generated by adding
    'change' values to the parent gene values.   The change values
    are derived from user judgements of good and bad.

o   Random marriage - in which child gene values are selected at
    random from the corresponding gene values of the parents.

o   In-between marriage - in which child gene values are generated by
    a randomly weighted average of parent gene values.

o   Spliced marriage - in which child gene values for the first n
    genes are taken from one parent, and the remaining values from
    the other parent.  n is selected randomly.  Splicing may be
    extended to more than two splice sections.

      The user is presented with several menu items to select between
the above algorithms.

      Disclosed is a solution in which the user has only three menus
item/icons to control the breeding:

o   Good - is used to judge a selected item as good.  Repeated
    selection can be made for very good etc.

o   Bad - is used to judge a selected item as bad.  Repeated
    selection can be made for very bad etc.  :Breed - is used to tell
    the computer when all judgements have been made, and to make a
    generation of children.

      The system decides automatically which of the breeding
algorithms to use based on the good/bad judgements.   Suppose for
example that one object is judged good and another very good.  The
system may choose to breed in one of several ways.

o   Use the judgements to define a steering direction, and...