Deduction reduction in expert systems

Question

In a system of rules, or some system of reasoning, which deduces facts through rules of inference in a straight chain, how would you cut off "unnecessary" branches? I'm not sure what formal terminology is, but I'm just trying to understand how people can limit their thoughts when they talk about problems, while all the semantic reasoning that I saw cannot do this.

For example, in an article by John McCarthy, An Example for Understanding the Natural Language and the Problems that it Raises , he describes potential problems in getting a program to intelligently answer questions about a news article in the New York Times. In section 4, “The Need for Nonmonotonic Reasoning,” he discusses the use of Occam Razer to limit the inclusion of facts in reasoning about history. A typical story that he uses concerns robbers who become victims of the owner of a furniture store.

If the program was asked to formulate a “minimal completion” of the story in predicate calculus, it may be necessary to include facts not mentioned directly in the original story. However, he will also need some way of knowing when to limit his deduction chain so as not to include irrelevant details. For example, he may want to include the exact number of police officers involved in the case, which is missing from the article, but he will not want to include the fact that every police officer has a mother.

Answer 1

Good question.

, , "", , ex ante, .. . "" - : post post, . ( ML "", .) , - " ", , ex post, () , , .

, , , - , , , , - Machine Learning; , ML - , (, SVM, - , .. " " ) ( , " ", , , , .. "", , , ).

?

w/r/t ML-, - , MARS - over-fit (a , , ( ). (DT, NN) (MARS) .

-, , /?

/? . " "; , ( , ) , , (, ). node , node, , node node. - () node.