Читать книгу Mivar NETs and logical inference with the linear complexity - Олег Варламов - Страница 4

Possibilities and limitations of production approach

Оглавление

Production approach possesses some important advantages. D.A. Pospelov proposes 9 types of productions and the possibility of other types to exist is underlined [23]. There are examples of knowledge represented in the form of productions. Pospelov defines production system as an aggregate of productions, which can include productions of all listed types. There are some production constructions. The general form is

i, П, Р, A=>B, Q

Here, A=>B is an ordinary production “if… else…” called a production core. P characterizes external condition or applicability conditions of production determined by factors which are not included in A. Condition P allows to chose needed productions from all production with A in left part of the core. П characterizes the sphere of the subject domain of the knowledge base, or pre-conditions of production applicability. These preconditions are not different from Р but they form the formal system in the frames of which logic reasoning will be drawn. Q characterizes post-conditions of the production indicating the changes that need to be brought into the base of knowledge and the system of productions after implementing of this production [23, p. 134-135].

In general form, productions are rather rare. Horoshevsky marks out an intermediate “layer of rules” for which the research of different logic inference on rules systems is made [31, p.82-83].

Logic processing is understood as some conclusion lying in the base of human reasoning. Each process of drawing a conclusion depends on the examination of options. To increase the efficiency of process of logic inference is a central problem of all systems of the deductive conclusion [23, p. 79].

D.A. Pospelov has described two methods of logic inference: the method of forward wave and the method of backward wave. In the first method, the wave of the searching paths is propagating from all initial areas to the aim area. In the second, this wave is propagating from the aim area towards initial areas. The difference of two methods comes to different number of steps of the search. The combined method is frequently used when forward and backward waves meet, and the path from initial data towards the aim is being driven [23, p.85]. Picture 1 shows an example the graph of non-oriented “AND-OR” net [23, p. 83].

Mivar NETs and logical inference with the linear complexity

Подняться наверх