Because we want to use engineered intelligent systems in the realm of computer science we have to map the engineering framework further into a computational framework. The basic concept of all computation is the automaton, which means in the full form a turing machine2.4. Every other kind of formalism in the realm of computer science is either a subconcept of a turing machine or it does not belong to computer science.
Figure 2.7:
Computational Framework For Intelligent Systems
|
Mapping the evolutionary and the engineering framework into a computational framework is rather straightforward (cf. figure 2.7).
- Automaton: The starting point is the concept of an automaton which consists of a finite table of commands telling how the read-write head has to be moved on the tape and how the automaton reacts to the input.
- Translation: To use the idea of biological systems based on genetic information one has to establish a translation algorithm
to map a genetic information into a working automaton.
- Population: A population then is given by a set of genetic informations
called gene pool as well as a set of automata
generated by a translation from the genetic informations given in the gene pool.
- Mating: During their life time will the members of the population occasionally directly exchange their genetic information through a mating operation
. The new genetic information is added to the gene pool. When a member terminates its process its genetic information will be deleted in the gene pool.
- World Interface: The original tape of the automaton is expanded to a multi tape representing an interface between the population and the assumed world .
- World: The world is a function 'behind' the interface which can read and write onto the interface. The world is either some part of the real world which is interfacing with the population or a virtual model which mimics some properties of the real world. In ideal cases the world can be stationary, but the usual case is a non-stationary world which is a mixture of partial regularities and noise based on many random-like processes.
Gerd Doeben-Henisch
2012-03-31