.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Down to page [NCSA Mosaic users only]
.
A variety of new computer techniques enable problems to be solved by successive improvements and developments. The simplest method consists of measuring a given performance and selecting the solutions which represent a small improvement whilst rejecting the remainder. This simple approach works for optimizing already satisfactory solutions, where only a small variation of quality is required. In cases where an already well-developed solution does not exist, or a possible improvement may be radically different, more sophisticated techniques have to be developed.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Genetic algorithms
Down to page [NCSA Mosaic users only]
.
Genetic algorithms
Genetic algorithms are a class of highly parallel evolutionary, adaptive search procedures. They are characterized by a string-like structure equivalent to the chromosomes of nature. These represent a coded form of parameters which control the problem being investigated. They are described as highly parallel because they search using populations of potential solutions rather than searching randomly or adjusting a single potential solution. Since optimum solutions are obtained by small, gradual changes within the population over several generations, they are defined as adaptive. Selection from the population occurs according to a measure of fitness criteria.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Criteria for selection
Down to page [NCSA Mosaic users only]
.
Criteria for selection
In our general model it is necessary to develop the coded version of the concept
to address a higher level of complexity. This is done by applying genetic
algorithms, classifiers or other evolutionary programming techniques.
The genetic code is bred into populations which are
developed into abstract models suitable for evaluation in a simulated
environment.
The criteria for selection must be carefully considered and
specified. The genetic code of the selected models is then used to breed further
populations in a cyclical manner. At any time abstract
models can be externalized for further examination or prototyping. This
externalization requires a further step of transformation or mapping.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Classifier system
Down to page [NCSA Mosaic users only]
.
Classifier system
Selection is controlled by a hierarchy of nested classifier systems, derived from
epigenetic algorithms - modified forms of a classifier algorithm. Normal crossover and mutation are used to breed the
populations.
The rules of a normal classifier system are logical, matching stimulus to action.
However, in our case, what actually occurs is more complex and closer to the
transition rules of a hierarchy of nested cellular automata: the effect is one of
environment response to stimulus. The response evaluated for the genetic
algorithm therefore determines the emergent properties of the system as a whole.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Artificial life
Down to page [NCSA Mosaic users only]
.
Artificial life
In nature it is only the genetically coded information of form which evolves, but
selection is based on the expression of this coded information in the outward
form of an organism. The codes are manufacturing instructions but their precise
expression is environmentally dependent.
Our architectural model, considered as a form of artificial life, will also
contain coded manufacturing instructions which are environmentally dependent, but
as with the actual model it is only the code script which evolves.
All the parts of the model co-operate and in that sense it can be considered as
an organism, but it will only fully exist as such if it is a member of an
evolving system of organisms interacting with each other as well as with the
environment.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Evolutionary techniques
Down to page [NCSA Mosaic users only]
.
Evolutionary techniques
In order that natural selection should work, certain criteria must be satisfied.
- The genetic information must replicate accurately.
- There must be the opportunity for the generation of variety and mutation (usually achieved by genetic crossover and very small random errors in the genetic copying).
- Any variation must also be capable of reproduction, and must occasionally confer potential advantage when expressed as a phenotype.
- There must be massive overproduction of phenotypes.
- There must be selective competition in particular environments (before replication of the genetic code).
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.