Page 140 - 49A Field Guide to Genetic Programming

126 12 Applications

Kell and his colleagues in Aberystwyth have had great success in applying
GP widely in bioinformatics (see infrared spectra above and (Allen, Davey,
Broadhurst, Heald, Rowland, Oliver, and Kell, 2003; Day, Kell, and Griffith,
2002; Gilbert, Goodacre, Woodward, and Kell, 1997; Goodacre and Gilbert,
1999; Jones, Young, Taylor, Kell, and Rowland, 1998; Kell, 2002a,b,c; Kell,
Darby, and Draper, 2001; Shaw, Winson, Woodward, McGovern, Davey,
Kaderbhai, Broadhurst, Gilbert, Taylor, Timmins, Goodacre, Kell, Alsberg,
and Rowland, 2000; Woodward, Gilbert, and Kell, 1999)). Another very
active group is that of Moore and his colleagues (Moore, Parker, Olsen, and
Aune, 2002; Motsinger, Lee, Mellick, and Ritchie, 2006; Ritchie, Motsinger,
Bush, Coffey, and Moore, 2007; Ritchie, White, Parker, Hahn, and Moore,
2003).
Computational chemistry is widely used in the drug industry. The prop-
erties of simple molecules can be calculated. However, the interactions be-
tween chemicals which might be used as drugs and medicinal targets within
the body are beyond exact calculation. Therefore, there is great interest in
the pharmaceutical industry in approximate in silico models which attempt
to predict either favourable or adverse interactions between proto-drugs and
biochemical molecules. Since these are computational models, they can be
applied very cheaply in advance of the manufacturing of chemicals, to decide
which of the myriad of chemicals might be worth further study. Potentially,
such models can make a huge impact both in terms of money and time
without being anywhere near 100% correct. Machine learning and GP have
both been tried. GP approaches include (Bains, Gilbert, Sviridenko, Gas-
con, Scoffin, Birchall, Harvey, and Caldwell, 2002; Barrett and Langdon,
2006; Buxton, Langdon, and Barrett, 2001; Felton, 2000; Globus, Lawton,
and Wipke, 1998; Goodacre, Vaidyanathan, Dunn, Harrigan, and Kell, 2004;
Harrigan et al., 2004; Hasan, Daugelat, Rao, and Schreiber, 2006; Krasno-
gor, 2004; Si, Wang, Zhang, Hu, and Fan, 2006; Venkatraman, Dalby, and
Yang, 2004; Weaver, 2004).

12.8 GP to Create Searchers and Solvers –
Hyper-heuristics

Hyper-heuristics could simply be defined as “heuristics to choose other
heuristics” (Burke, Kendall, Newall, Hart, Ross, and Schulenburg, 2003).
A heuristic is considered as a rule-of-thumb or “educated guess” that re-
duces the search required to find a solution. The difference between meta-
heuristics and hyper-heuristics is that the former operate directly on the
problem search space with the goal of finding optimal or near-optimal so-
lutions. The latter, instead, operate on the heuristics search space (which
consists of the heuristics used to solve the target problem). The goal then