Page 142 - 49A Field Guide to Genetic Programming

128 12 Applications

The use of GP in computer art can be traced back at least to the work
1
of Sims (Sims, 1991) and Latham. Jacob’s work (Jacob, 2000, 2001) pro-
vides many examples. McCormack (2006) considers the recent state of play
in evolutionary art and music. Many recent techniques are described in
(Machado and Romero, 2008).
Evolutionary music (Todd and Werner, 1999) has been dominated by
Jazz (Spector and Alpern, 1994). An exception is Bach (Federman, Spark-
man, and Watt, 1999). Most approaches to evolving music have made at
least some use of interactive evolution (Takagi, 2001) in which the fitness
of programs is provided by users, often via the Internet (Ando, Dahlsted,
Nordahl, and Iba, 2007; Chao and Forrest, 2003). The limitation is al-
most always finding enough people willing to participate (Langdon, 2004).
Costelloe and Ryan (2007) tried to reduce the human burden. Algorithmic
approaches are also possible (Cilibrasi, Vitanyi, and de Wolf, 2004; Inagaki,
2002).
One of the sorrows of AI is that as soon as it works it stops being AI (and
celebrated as such) and becomes computer engineering. For example, the
use of computer generated images has recently become cost effective and is
widely used in Hollywood. One of the standard state-of-the-art techniques
is the use of Reynold’s swarming “boids” (Reynolds, 1987) to create ani-
mations of large numbers of rapidly moving animals. This was first used in
Cliffhanger (1993) to animate a cloud of bats. Its use is now commonplace
(herds of wildebeest, schooling fish, and even large crowds of people). In
1997 Reynold was awarded an Oscar.
Since 2003, EvoMUSART (the European Workshop on Evolutionary Mu-
sic and Art) has been held every year along with the EuroGP conference as
part of the EvoStar event.

12.11 Compression

Koza (1992) was the first to use genetic programming to perform compres-
sion. He considered, in particular, the lossy compression of images. The idea
was to treat an image as a function of two variables (the row and column
of each pixel) and to use GP to evolve a function that matches as closely as
possible the original. One can then use the evolved GP tree as a lossy com-
pressed version of the image, since it is possible to obtain the original image
by evaluating the program at each row-column pair of interest. The tech-
nique, which was termed programmatic compression, was tested on one small
synthetic image with good success. Programmatic compression was further
developed and applied to realistic data (images and sounds) by Nordin and
Banzhaf (1996).

1
http://www.williamlatham1.com/