Page 249 - 49A Field Guide to Genetic Programming

Colophon

A
This book was primarily written using the LT X document preparation
E
system, along with BibT X, pdflatex and makeindex. Most of the editing
E
was done using the emacs and xemacs editors, along with extensions such
as RefT X; some was done with T XShop as well. Most of the data plots
E
E
were generated using gnuplot and the R statistics package. Diagrams were
generated with a variety of tools, including the Graphviz package, tgif and
xfig. A whole host of programming and scripting languages were used to
automate various processes in both the initial scientific research and in the
production of this book; they are too numerous to list here, but were crucial
1
nonetheless. The cover was created with Adobe Photoshop and gimp.
Coordinating the work of three busy, opinionated authors is not trivial,
and would have been much more difficult without the use of revision control
systems such as Subversion. Around 500 commits were made in a six month
period, averaging around 10 commits per day in the final weeks. The actual
files were hosted as a project at http://assembla.com; we didn’t realise
until several months into the project that Assembla’s president is in fact
Andy Singleton, who did some cool early work in GP in the mid-90’s.
The “reviews” and “summaries” on the back cover were generated
stochastically using the idea of N-grams from linguistics. For the “reviews”
we collected a number of reviews of previous books on GP and EAs, and
tabulated the frequency of different triples of adjacent words. These fre-
quencies of triples in the source text were then used to guide the choices of
words in the generated “reviews”. The only word following the pair “ad”
and “hoc” in our source reviews, for example, was “tweaks”; thus once “ad”
and “hoc” had been chosen, the next word had to be “tweaks”. The pair
“of the”, on the other hand, appears numerous times in our source text, fol-
lowed by words such as “field”, “body”, and “rapidly”. However, “theory”
is the most common successor, and, therefore, the most likely to be cho-
sen to follow “of the” in the generation of new text. The generation of the
“summaries” was similar, but based on the front matter of the book itself.
See (Poli and McPhee, 2008a) for an application of these ideas in genetic
programming.

1 Adobe Photoshop is a registered trademark of Adobe Systems Incorporated