Catalogue of Artificial Intelligence Techniques


Jump to: Top | Entry | References | Comments

View Maths as: Images | MathML

Control Knowledge

Aliases: Search Control Knowledge or Meta Knowledge

Categories: Knowledge Representation

Author(s): Guowei Deng

Control Knowledge In large knowledge bases that contain a lot of rules, the intractability of search has to be concerned. When there are many possible paths of reasoning, it is very important that useless ones are not be taken. Knowledge about which paths are most likely to lead quickly to a goal state is normally called Control Knowledge( Search Control knowledge or meta­knowledge).There are several forms of control knowledge: 1.Knowledge about which states are more preferable to others. 2.Knowledge about which rule to apply in a given situation. 3.Knowledge about the order in which to pursue sub­goals. 4.Knowledge about useful sequences of rules to apply. A number of AI systems represent their knowledge with control rules. SOAR and PRODIGY are two of these AI systems. SOAR is a general architecture for building intelligent systems. It is based on a set of specific, cognitively motivated hypotheses about the structure of human problem solving. These hypotheses are derived from what we know about short­term memory, practice effects, etc. In SOAR, Long­term memory is stored as a set of productions(rules).Short­term memory (working memory) is a buffer that is affected by perceptions and serves as a storage area for facts deduced by rules in long memory. It is similar with the state description in problem solving. All problem­ solving activities take place as state space traversal and all intermediate and final results of problem solving are remembered for future reference. In figure1,it shows what the control rules look like: Under conditions A and B, rules that do {not} mention X {at all, in their left­hand side, in their right­hand side} will {definitely be useless, probably be useless ... probably be especially useful definitely be especially useful} PRODIGY is a general­purpose problem solving system that incorporates several different learning mechanisms. A good deal of the learning in PRODIGY is directed at automatically constructing a set of control rules to improve search in a particular domain. It can acquire control rules in a number of ways: *Through hand coding by programmers. *Through a static analysis of the domain's operators. *Through looking at traces of its own problem­solving behavior. It learn control rules from both its experience and failures. If PRODIGY pursues an unfruitful path, it will try to come up with an explanation of why that path failed. It will then use that explanation to build control knowledge that will help it avoid fruitless search paths in the future. Above all, A successful AI system always has a series of control rules that can represent its control knowledge rationally and successfully.



Add Comment

No comments.