Geoscience Reference
In-Depth Information
by three simple case studies are presented with the intention of inspiring readers to develop and
test their own ideas about what GP can be used for. Finally, a short discussion on the future direc-
tions of GP helps steer some searching philosophical questions for the reader to consider. One of
the key messages from the first edition version of this chapter remains suitably apt - 'if in doubt,
experiment'! It is hoped that both newcomers and experienced users will use this chapter to enhance
their understanding of GP and at the same time recognise and absorb what is on offer as a powerful,
rewarding, data-driven modelling tool.
8.1 KEY FACTS ON GENETIC PROGRAMMING
You have to kiss a lot of frogs before you find your handsome prince
Popular Saying*
This chapter has been written for aspiring practitioners with little or no knowledge of genetic pro-
gramming (GP) but with a strong interest in using the technique to discover solutions to difficult
geographical problems. It provides an outline of the underlying concepts of GP which is reinforced
by a description of a general framework for the methodological stages that are associated with its
implementation. Three case studies are presented to give an idea of the kind of applications that
GP is demonstrably good at (1) estimating pan evaporation estimation at Elephant Butte Reservoir,
United States; (2) rainfall-runoff modelling of the Annapolis River, Canada; and (3) building a
spatial interaction model for County Durham, United Kingdom. The chapter concludes with a short
discussion which is characterised by some searching, potentially contentious, philosophical ques-
tions about the future of GP. In doing so, it builds on earlier prophecies made in the first edition of
this topic. Interested readers would beneit from re-reading that material (Diplock, 2000).
From a historical perspective, vast technological leaps in computing have been made since the
original version of this chapter was written and published. Computers are faster and more com-
pact, modelling takes minutes or hours, rather than days or weeks, and the advent of proprietary
software means that users are freed from the arduous and perhaps largely off-putting task of soft-
ware development and coding. Other tools which have sped up modelling operations are multiple-
core and cloud-based processing technologies, giving widespread access to much faster and more
powerful evolutionary options. The reader is also encouraged to take advantage of a complemen-
tary chapter contained in this second edition, on the potential benefits of parallel computing (PC)
(Adnan et al., 2014) approaches for performing GeoComputation (GC). Another huge improvement
is the availability of specialist user-friendly software packages, such as
Discipulus
(http://rmltech.
com/),
Eureqa
(http://creativemachines.cornell.edu/eureqa/) and
GeneXproTools
(http://www.
gepsoft.com/), each of which avoids the need to design or code your own GP program. These or
similar packages have low-cost, free or demonstration versions which permit both experimentation
and familiarisation with particular software environments. For those with a strong preference for
command-line programming, options exist for the implementation of GP in
R
(http://r-project.org/)
and MATLAB
®
(http://mathworks.com/), in addition to more traditional programming languages
such as Fortran.
GP is all about breeding equations. Simply put, it works by applying an algorithm that is itself
embedded into a piece of software, which, when run, concurrently delivers multiple solutions
fitted to a particular dataset. These equations are checked automatically by the software to see
how well they perform. Then, small quasi-random alterations are made to create a new set of
solutions, whereafter performance is again checked. This is an iterative process that continues
until it is brought to an end by the user. The Frog King analogy at the start of this chapter is
apposite because to evolve a solution that the user is happy with requires a lot of patience and
*
In the original
Frog King
(Grimm and Grimm, 1812) fairy tale, a spoiled princess reluctantly befriends a frog. In modern
versions of that tale, the frog is magically transformed into a handsome prince after the princess kisses it.
