Adaptation
Adaptation is a response to environmental conditions: the process of adjustment to given (outer) conditions. It refers to a any change in the structure or function of an entity (say, a biological organism or a software system) that allows it to act more effectively, efficiently and successfully in its environment. In general it is the ability to learn from experience and describes the process of change and modification in a system that helps to adjust and fit the system to the conditions of the environment. Adaptation is the set of adjustments made by systems in respect of their environments. The adjustments are often controlled by experience and practice. The concept of a change in behavior or structure as a result of experience and practice is related to learning. Learning is a special form of adaptation for the behavior of intelligent systems.
Function, Purpose and Niche Occupation
Typical for adaptive systems is the existence of a function or purpose: they occupy a niche and play a role, specified by the context of the environment. A snowflake is a beautiful crystal with a complex structure, but it has no particular function or purpose. Biological organisms affected and shaped by evolution have nearly always a function in a larger ecosystem, even the organs of an organism have a particular function. The organs, especially the sense organis, can be adaptive itself. In physiology, adaptation is the responsive adjustment of a sense organ (as the eye) to varying conditions (as of light).
In evolutionary biology, adaptation is a physiological process or behavioral trait of an organism that has evolved over a period of time, because it increases the expected long-term reproductive success of the organism. A individual in an evolutionary system adapts to survive: it changes in order to persist. This is related to the Red Queen Effect: a situation in nature where individuals must adapt constantly to changing environments just to survive.
Adaptation occurs in evolution often through natural selection at the population level. It is the process of accommodation and adjustment to an ecological or economic niche. While adaptation is the process of being optimized for some function, exaptation is "meta-adaptation": the adaptation of an existing trait or property for a new function. Exaptation characterizes components, traits and properties which arose as the result of adaptation for one function, and which have later been used for a new function. Though developed for a particular purpose, an organism may eventually use a structure or attribute for a different, or exaptive, purpose.
Optimization of Fitness
Individuals try to occupy their niche in an optimal way, or a driven by evolutionary forces and natural selection to do it. The optimal occupation of a niche is at the maximum (or minimum) of the fitness function. Since optimization means maximization (or minimization) of a certain function, adaptation in evolution is therefore nothing else but optimization of fitness, and optimal occupation of the corresponding niche.
Optimization methods are besides Evolutionary Algorithms for example Gradient Descent (used in artificial neural networks and "back propagation"), Tabu Search and Simulated Annealing.
Requirements
A necessary condition for adaptation is flexibility or the ability for reconfiguration. Only a system with a flexible structure is able to evolve into something new. Decentralized and distributed systems with high redundancy and scalability (the ability to growth) are usually more flexible than centralized systems.
The law of indispensable or requisite variety from William Ross Ashby states simply that only variety in a system itself can successfully counter a variety of disturbances in the environment: a wide variety of available responses and actions is indispensable in order to ensure that a system which aims to maintain itself in a certain state can actually adapt itself satisfactorily if it is confronted with a wide variety of pertubations from the outside.
This may seem obvious, because a flexible system with many options is of course better able to cope with change and changing conditions. In other words, "the larger the variety of actions available to a control system, the larger the variety of perturbations it is able to compensate" see [1].
It is also clear that sufficient "requisite variety" is already available in systems with a small numbers of elements, as soon as those elements can interact in arbitrary ways we get a combinatorial explosion. Thus the law might say nothing, but nevertheless there is some truth in it.
Examples
Adaptation is typical for biological and biologically inspired systems as neural networks. It is a property of many intelligent systems. Intelligent agents, neural networks and adaptive systems are examples of adaptive systems: intelligent agents adapt their behavior, neural networks their connections and weights, and complex adaptive systems in general their components, relations and schemes. As Pattie Maes mentions in [Maes (1994)], "Adaptive means that the agent improves its goal-achieving competence over time."
The time-scale of adaptation can vary massively: in nervous systems and neural networks from seconds to hours, in the immune system hours to days, in business firms months to years, in species days to centuries, and in ecosystems years to millennia (see Holland's Book Hidden Order: How Adaptation Builds Complexity).
A number of physiological, psychological and sociological phenomena can be considered as adaptation:
Relations
Adaptation is related to homeostasis, the process of maintaining a system at a constant, stable state in a changing and instable environment. It allows a system to maintain consistent and suitable behavior despite clearly visible variations and differences in different operating contexts and environments. Homeostasis is based on regulatory and feedback mechanisms, whereas adaptation is more related to structural adjustments of the system.
The process of adaptation is often connected to an increase in complexity and acquisition of information from the environment or in other words to learning, but also to specialization and differentiation.
Adaptation is also related to optimization, since an organism adapted to a certain (ecological) niche is optimal suited to fulfil the corresponding task most efficiently.
Articles
- Modeling Adaptive Autonomous Agents, Pattie Maes (1994)
- Richard B. Mazess, Biological Adaptation: Aptitudes and Acclimatization , in Watts, E.S., F.E. Johnston, and G.W. Lasker. 1975. Biosocial Interrelations in Population Adaptation. The Hague:MoutonPublishers. pp. 9-18.
Books
Lashon Booker (Editor), Perspectives on Adaptation in Natural and Artificial Systems (Proceedings Volume in the Santa Fe Institute Studies in the Sciences of Complexity), Oxford Univ Press (2005) ISBN 0195162935
John H. Holland, Hidden Order: How Adaptation Builds Complexity, Addison Wesley Publishing Company, 1996, ISBN 0201442302
John H. Holland, Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence, The MIT Press, 1992, ISBN 0262581116
