HollandAdaptationCover.jpg

A book originally from 1975, reprinted in 1992 with an additional chapter.

The work proposes a formal framework for studying adaptivity, and shows how the genetic processes can be generalized to be useful in many different disciplines (the 1975 book is the one that reportedly brought genetic algorithms into wider use). The approach indeed seems to be not presenting genetic evolution as just an example of an adaptive process, but as a (or The) source of inspiration and ideas for understanding all kinds of adaptive systems.

A very short summary of the formal framework:

An adaptive system could be specified by a set of objects (A, Omega, I, tau), where:

  • A is the set of attainable structures (the domain of action of the adaptive plan).
  • Omega is the set of operators for modifying structures.
  • I is the set of possible inputs to the system from the environment.
  • tau is the adaptive plan which, on the basis of the input and structure at time t, determines what operator is to be applied at time t.

A problem in adaptation would be said to be well posed once the following constructs have been specified:

  • A set of plans, given either explicitly or implicitly, which are candidates for comparison.
  • A set of possible environments (or states of the environment) that the system may encounter.
  • A criterion for comparing the efficiency of different plans (under the uncertainty represented by the set of possible environments).

Reading the book was quite useful for me, as I am trying to understand the concept of adaptivity. But if your goal is to understand genetic algorithms, I'm sure you can find more up to date and easier to follow sources for that.

More information about the book at MIT Press: http://mitpress.mit.edu/catalog/item/default.asp?ttype=2&tid=7593

and at Amazon: http://www.amazon.com/dp/0262581116