JPH02278338A - Table-type backward inference system - Google Patents

Abstract

PURPOSE:To attain backward inference by defining procedual contents dealing a numerical formula relating squares in a table and non-numerical data, and inferring an unknown number from the value of a right side in the formula and the past rule of thumb in table-type inference. CONSTITUTION:An inference engine 2 executes inference by using the content of a knowledge base 1 in which the formula and the rule mentioned between cells are accumulated, and they are displayed in a table-type interface 3. There are a forward inference part and a backward inference part in the inference engine 2, and they control forward inference and backward inference. In the allocation of the numerical formula, the value inputted to the right side is allocated to whole unknown variables based on previously inputted values. In the rule, whole matters are retrieved by using value candidates which have previously been set, all the values of the unknown variables in a left side, which fits to the value of the variable in the right-side, whereby the value detected first is shown. Thus, the decision of a will can efficiently be supported in the case of trial error so that the variables of the whole variables are adjusted to the value of the variable in the right side by variously changing them.

Description

[Detailed Description of the Invention] [Field of Industrial Application] This invention relates to backward inference in two-table inference.

[Conventional technology]

For example, the conventional [TK/SolwerJ (Univac pamphlet)] has a tabular interface and uses a formula to relate the squares in nine tables, ``Variable a, 10 variables, Raw variable C = Variable d''. The variables a, b on the left side are
Assuming that we have defined a formula in which the variable d on the right side is determined by , c, +41. (51, (61 are the variables a, b, c on the left side, (7) is the variable d on the right side, and the three variables on the left side each have the value "I" @ 2
II. Substituting "3°" will be calculated and "5" will be displayed in the variable d.Also, even if the variable C on the left side is unknown and you want to find this value, there is no need to transform the equation, and the variable d on the right side By inputting the value "10" into the field, as shown in Figure 6, the already defined formula can be inversely calculated to obtain the value "7" of the variable C and displayed in 2. However, if the unknown variable is two If there are more than one, inverse calculation is not possible.

[Problem to be solved by the invention]

Conventional tabular inference methods work as described above, so when there is one unknown variable in a mathematical formula, it is possible to perform inverse calculations, but when there are two or more unknown variables, it is possible to perform inverse calculations with the defined formula. It is not possible to obtain a value in the direction. Therefore, for example, when the total value increases, if one variable is corrected as an unknown variable, a contradiction will not occur, but if the total change is absorbed by one variable, then it is not assigned to multiple variables I can't.

(1) Therefore, a great deal of trial and error is required to set the values of all variables to appropriate values. A problem arises. (2
) and procedural content that handles non-numeric values (hereinafter referred to as rules)
Since it is not possible to define or perform the opposite direction (warifuri), there are two issues: it is not possible to support decision-making other than numerical calculations.

This method solves the two problems mentioned above, and even if there are unknown variables on two Juls in the defined formula or rule, it can be reversed using the values input so far. The purpose is to be able to reason.

[Means to solve the problem]

The tabular backward inference method according to the present invention infers in a defined direction.I+? In contrast to directional inference, a mode is provided in which only inference is performed in the opposite direction.

In this mode, in the case of a mathematical formula, the value given on the right side can be distributed to all variables on the left side, or assigned to a few specific variables, referring to the value input before inference. In the case of rules, all candidates are searched using the candidate values of the already defined values, and the method that matches the value of the used variable is used. It is a means of determining the variables to be used (hereinafter referred to as "rule substitution").

[Effect]

In this invention, when changing a mathematical formula, it is proportional to the value input so far.

Distribute the value input on the right side to all unknown variables using either the equality method or the ratio method, and in the rule,
A complete search is performed using the already set value candidates to find all variables on the left side that match the value of the variable on the right side, and the first one found is shown. However, this applies to all variables on the left side, and if there is a variable whose value should be saved, the value of that variable is saved by setting it as "fixed."

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram of the tabular inference method according to the present invention, in which (1) is the knowledge base, (2) is the inference engine, (
3) is a tabular interface.

A knowledge base (1
), the inference engine (2) makes an inference and displays it on the tabular interface (3). Inference engine (2)
has a j1 direction reasoning section and an i direction reasoning section, ilk! 1
Control both directional and backward inference.

As shown in FIG. 2, which has the same reference numerals as those in FIG. 5, in the forward mode, the operation is the same as in FIG. When inputting , the variable a and f number b-i1! are set to match the value of fed as shown in Figure 3. The value of Ic is changed. Said sixth
If you want to change only IJ of variable C as shown in the figure,
This can be realized by fixing the values of f-number a and f-number S outside Ll.

Figure 4 is a flowchart in backward inference mode. First, when you input a value into the variable cell on the right side using 5(4Il), use sha to determine whether what is defined in that cell is a formula or a rule. If formula 9 is defined, first S
Select the distribution method using ρj. There are three methods for distributing the change in the value of the variable cell on the right side: "proportional distribution,""equaldistribution," and "ratio distribution." However, if the variable cell on the left side already has a value If it is not input, it is "equal", and if it is input and backward inference has already been performed, use the same method as 111, and if it is the first time. Select "Proportional". Using this method, 5 (4
In step 4, if there is a value input up to that point, change the value using various methods such as referring to it. On the other hand, in the case of rules, it is not possible to perform reverse execution using heuristics like in the case of mathematical expressions, so S(t)
Using other candidates for the variables on the left side that have already been set,
In step 8 (46), all combinations are executed in 111 directions to find one that matches the value of the input variable on the right side, and the first one found in the S envelope η is displayed.

In addition, 15tep regarding the backward inference method in table format.
Although the example shown in FIG.

In addition, in the above embodiment, backward inference in a tabular format in which mathematical formulas and rules can be written has been explained. A similar effect can be achieved in systems.

Further, in the above embodiment, the value candidates are used for backward inference of rules, but the same effect can be obtained even if they are used for backward inference of mathematical expressions.

〔Effect of the invention〕

As described above, according to the present invention, the values of all variables can be determined to match the value of the variable on the right side that should be determined by the formula and the rule, so when the value of the variable on the right side changes slightly, etc. This has the effect of being able to efficiently support decision-making in cases where trial and error is performed by changing the values of all variables to match the values of the variables on the right-hand side.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the present invention, Fig. 2 is a diagram showing an example of forward inference and an operation example for implementing backward inference, and Fig. 3 is a diagram showing an example of this invention (backward inference). 4 is a flowchart of the present invention, FIG. 5 is a diagram showing an example of forward inference and an operation example for implementing conventional inverse calculation, and FIG.
The figure shows an example of conventional inverse calculation. In the figure, (1) is the knowledge base, (2) is the inference engine (
3) is a tabular interface.

Claims (1)

[Claims] In tabular reasoning, there is a means for defining mathematical formulas that handle numerical data that relate between squares in the table, and procedural content that handles non-numeric data, and all values on the left side of the formula. A tabular backward inference method characterized by having means for inferring an unknown quantity from the value on the left side and past empirical rules even if it is not determined.