README
1+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
2
3Co-evolutionary Symbolic Regression (coev_symbreg):
4Mixed real-valued GA - GP co-evolution example with Open BEAGLE
5
6Copyright (C) 2003
7by Jiachuan Wang <jiacwang@ecs.umass.edu>
8and Christian Gagne <cgagne@gmail.com>
9
10+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
11
12
13Getting started
14===============
15
16 Example is compiled in binary 'coev_symbreg'. Configuration file for
17 real-valued GA population (the evolving training set) is in file
18 'trainset-thread.conf', while configuration file for GP symbolic
19 regression population is in file 'symgp-thread.conf'.
20
21Objective
22=========
23
24 Two populations competitive co-evolution for symbolic regression.
25 First population is made of standard GP mathematical expressions.
26 Second population is made of individual representing 20 $(x_i,y_i)$
27 samples in the domain $[-1,1]$ of the equation $x^4+x^3+x^2+x$.
28 The global objective is to a find symbolic expressions that
29 "generalizes" well over the domain sampled.
30
31Terminal set of GP expressions
32==============================
33
34 X (the independent variable)
35 Ephemeral constants randomly generated in $[-1,1]$
36
37Function set of GP expressions
38==============================
39
40 +
41 -
42 *
43 / (protected division)
44
45Representation of training sets
46===============================
47
48Vector of 20 real-valued numbers in $[-1,1]$.
49
50Fitness
51=======
52
53 Root Mean Square (RMS) error of the GP expressions on the 20 samples
54 of the training set. GP expression minimizes the RMS error, while
55 training set maximizes it. Individuals of one population are
56 evaluated against last generation other population best performing
57 individual.
58
59Reference
60=========
61
62Liviu Panait and Sean Luke, Methods for Evolving Robust Programs,
63Proceedings of Genetic and Evolutionary Computation -- GECCO-2003,
64LNCS, Vol. 2724, pp. 1740-1751, Springer-Verlag, 12-16 July 2003.
65
66