"A useful general definition of mental models must capture several features inherent in our informal descriptions. First, a model must make it possible for the system to generate predictions even though knowledge of the environment is incomplete. Second, it must be easy to refine the model as additional information is acquired without losing useful information already incorporated. Finally, the model must not make requirements on the cognitive system's processing capabilities that are infeasible computationally. In order to be parsi- monious, it must make extensive use of categorization, dividing the environment up into equivalence classes." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"Although mental models are based in part on static prior knowl- edge, they are themselves transient, dynamic representations of par- ticular unique situations. They exist only implicitly, corresponding to the organized, multifaceted description of the current situation and the expectations that flow from it. Despite their inherently transitory nature - indeed because of it - mental models are the major source of inductive change in long-term knowledge structures." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"Classifier systems are a kind of rule-based system with general mechanisms for processing rules in parallel, for adaptive generation of new rules, and for testing the effectiveness of existing rules. These mechanisms make possible performance and learning without the "brittleness" characteristic of most expert systems in AI." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"Deduction is typically distinguished from induction by the fact that only for the former is the truth of an inference guaranteed by the truth of the premises on which it is based. The fact that an inference is a valid deduction, however, is no guarantee that it is of the slightest interest." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"How can a cognitive system process environmental input and stored knowledge so as to benefit from experience? More specific versions of this question include the following: How can a system organize its experience so that it has some basis for action even in unfamiliar situations? How can a system determine that rules in its knowledge base are inadequate? How can it generate plausible new rules to replace the inadequate ones? How can it refine rules that are useful but non-optimal? How can it use metaphor and analogy to transfer information and procedures from one domain to another?" (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"Our approach assumes that the central problem of induction is to specify processing constraints that will ensure that the inferences drawn by a cognitive system will tend to be plausible and relevant to the system's goals. Which inductions should be characterized as plausible can be determined only with reference to the current knowledge of the system. Induction is thus highly context dependent, being guided by prior knowledge activated in particular situations that confront the system as it seeks to achieve its goals. The study of induction, then, is the study of how knowledge is modified through its use." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"We will treat problem solving as a process of search through a state space. A problem is defined by an initial state, one or more goal states to be reached, a set of operators that can transform one state into another, and constraints that an acceptable solution must meet. Problem-solving methods are procedures for selecting an appropriate sequence of operators that will succeed in transforming the initial state into a goal state through a series of steps." (John H Holland et al, "Induction: Processes Of Inference, Learning, And Discovery", 1986)
"An internal model allows a system to look ahead to the future consequences of current actions, without actually committing itself to those actions. In particular, the system can avoid acts that would set it irretrievably down some road to future disaster ('stepping off a cliff'). Less dramatically, but equally important, the model enables the agent to make current 'stage-setting' moves that set up later moves that are obviously advantageous. The very essence of a competitive advantage, whether it be in chess or economics, is the discovery and execution of stage-setting moves." (John H Holland, 1992)
"Because the individual parts of a complex adaptive system are continually revising their ('conditioned') rules for interaction, each part is embedded in perpetually novel surroundings (the changing behavior of the other parts). As a result, the aggregate behavior of the system is usually far from optimal, if indeed optimality can even be defined for the system as a whole. For this reason, standard theories in physics, economics, and elsewhere, are of little help because they concentrate on optimal end-points, whereas complex adaptive systems 'never get there'. They continue to evolve, and they steadily exhibit new forms of emergent behavior." (John H Holland, "Complex Adaptive Systems", Daedalus Vol. 121 (1), 1992)
"The systems' basic components are treated as sets of rules. The systems rely on three key mechanisms: parallelism, competition, and recombination. Parallelism permits the system to use individual rules as building blocks, activating sets of rules to describe and act upon the changing situations. Competition allows the system to marshal its rules as the situation demands, providing flexibility and transfer of experience. This is vital in realistic environments, where the agent receives a torrent of information, most of it irrelevant to current decisions. The procedures for adaptation - credit assignment and rule discovery - extract useful, repeatable events from this torrent, incorporating them as new building blocks. Recombination plays a key role in the discovery process, generating plausible new rules from parts of tested rules. It implements the heuristic that building blocks useful in the past will prove useful in new, similar contexts." (John H Holland, "Complex Adaptive Systems", Daedalus Vol. 121 (1), 1992)
"Even though these complex systems differ in detail, the question of coherence under change is the central enigma for each." (John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"If we are to understand the interactions of a large number of agents, we must first be able to describe the capabilities of individual agents." (John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"Model building is the art of selecting those aspects of a process that are relevant to the question being asked. As with any art, this selection is guided by taste, elegance, and metaphor; it is a matter of induction, rather than deduction. High science depends on this art." (John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"[…] nonlinear interactions almost always make the behavior of the aggregate more complicated than would be predicted by summing or averaging." (John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"The multiplier effect is a major feature of networks and flows. It arises regardless of the particular nature of the resource, be it goods, money, or messages." (John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"With theory, we can separate fundamental characteristics from fascinating idiosyncrasies and incidental features. Theory supplies landmarks and guideposts, and we begin to know what to observe and where to act."(John H Holland, "Hidden Order: How Adaptation Builds Complexity", 1995)
"It may not be obvious at first, but the study of emergence and model-building go hand in hand. The essence of model-building is shearing away detail to get at essential elements. A model, by concentrating on selected aspects of the world, makes possible the prediction and planning that reveal new possibilities. That is exactly the problem we face in trying to develop a scientific understanding of emergence." (John H Holland, "Emergence", Philosophica 59, 1997)
"Shearing away detail is the very essence of model building. Whatever else we require, a model must be simpler than the thing modeled. In certain kinds of fiction, a model that is identical with the thing modeled provides an interesting device, but it never happens in reality. Even with virtual reality, which may come close to this literary identity one day, the underlying model obeys laws which have a compact description in the computer - a description that generates the details of the artificial world." (John H Holland, "Emergence", Philosophica 59, 1997)
No comments:
Post a Comment