JPH0465719A - Approximate inference device - Google Patents

Abstract

PURPOSE:To input a language value by providing the approximate inference device with a means for forming and storing data indicating a membership function corresponding to the set language value, and a language value interpolating means for forming data indicating the membership function of an input language. CONSTITUTION:A membership function forming device 32 synthesizes the knowledge of plural experts stored in a set language value storing device 31 and forms a membership function. The language value interpolating device 33 executes language value interpolating processing for the representative value of language values applied from a language value input device 14 and applies formed language value data to the device 14. When the language value applied from the device 14 coincides with a value already synthesized by the device 32, interpolation operation is unnecessary and language value data stored in the device 32 are directly read out. Consequently, a language value to be sensuous data such as heavy or light sense can be inputted.

Description

[Detailed Description of the Invention] Summary of the Invention By statistically processing the quantitative meanings of some set language values given by a plurality of experts, data representing a membership function corresponding to the language values is provided. Create and memorize it in advance. When an input language value other than the above set language value is given by data that specifies it, interpolation processing is performed on the above specified data using data representing the membership function of the above set language value, and the above input language value is Create data representing the membership function.

Further, the static information amount and clarity of the event are calculated using the membership function obtained by the interpolation process.

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an approximate reasoning device capable of inputting linguistic values.

Here, the term "linguistic value" refers to a set of values expressed through sensory language that defines vague quantitative ranges such as large, very large, small, heavy, light, and very light.

Prior Art Approximate inference is known in which the inference result is modified or changed depending on the amount of information about the event used to derive the inference result (for example, ZhangHongmin
AN EXPERT SYSTEM WITHHI
NKING INI MACES'', Preprin
ts of 5econd IFSA Congress
s.

Tokyo, July 20-25.1987. P
, 765).

This approximate inference method calculates the amount of information for each event (i.e., information identification ability of the event) using a membership function given to each conclusion for the event, and calculates the inference result (i.e., the probability that the conclusion holds true). The aim is to improve the discriminative ability of inference results by modifying or changing the information based on the amount of information of the event used to draw the conclusion (by multiplying the probability and the amount of information).

Such conventional approximate inference devices can only accept as input a definite value (referred to as a definite value) expressed by a single number. It was not possible to input verbal values such as sensory data such as heavy or light.

SUMMARY OF THE INVENTION OBJECTS OF THE INVENTION It is an object of the present invention to provide an approximate reasoning device capable of inputting linguistic values.

Structure of the Invention 1 Operation and Effect The approximate inference device according to the present invention statistically processes data given by a plurality of experts indicating the meaning range of each of a plurality of setting language values related to an event. means for creating and storing data representing a membership function corresponding to each setting language value; The present invention is characterized by comprising a language value interpolation means for performing interpolation processing on the specific data using the above-mentioned language to create data representing a membership function of input language values.

The approximate inference device further applies data representing a membership function created by interpolation processing regarding data for specifying input linguistic values to preset expert knowledge representing the relationship between an event and a conclusion. This provides an approximate reasoning means for inferring the possibility of a conclusion.

The approximate inference device further changes the data for specifying the linguistic value within the range width of the event at regular intervals, and the data and event representing the membership function created by the linguistic value interpolation means for each specific data. It is provided with static information amount calculation means for calculating the amount of linguistic value static information for each event using expert knowledge representing the relationship with the conclusion.

The above approximate inference device further includes a clarity calculation means that calculates the clarity of each event for each conclusion using the calculated amount of static information, and a clarity calculation means that uses the clarity obtained from the clarity calculation means to actually calculate the A clarity adding means is provided for calculating the clarity of an event whose value has been input.

The approximate inference device according to the present invention corresponds to each set language value by statistically processing data given by a plurality of experts indicating the meaning range of each of a plurality of set language values regarding an event. means for creating and storing data representing a membership function; a linguistic value input means for inputting data specifying a linguistic value to be input; and in response to specific data input to the linguistic value input means; Language value interpolation means is provided for performing interpolation processing on the specific data using data representing the membership function of the set language value to create data representing the membership function of the input language value.

The approximate inference device according to the present invention corresponds to each set language value by statistically processing data given by a plurality of experts indicating the meaning range of each of a plurality of set language values regarding an event. A means for creating and storing data representing a membership function; when data for specifying a language value is given, interpolation processing is performed on the specified data using data representing the membership function of the set language value; (1) linguistic value interpolation means for manually creating data representing a membership function of linguistic values, and data for specifying linguistic values within a range of an event, which are changed at regular intervals and provided to the linguistic value interpolation means; A static system that calculates the amount of linguistic value static information for each event using data representing membership functions created by the linguistic value interpolation means for each specific data and expert knowledge representing the relationship between events and conclusions. It is equipped with a means for calculating the amount of information.

The approximate inference device further uses a clarity calculation means that calculates the clarity of each event for each conclusion using the calculated amount of static information, and the clarity obtained from the clarity calculation means,
It is provided with a clarity adding means for calculating the clarity of an event in which a linguistic value is actually input.

According to this invention, it is possible to input linguistic values representing perceptual data such as large, small, heavy, light, etc., and approximate inference can be made using the input linguistic values. Furthermore, a membership function for any input language value is created through interpolation using data representing membership functions for preset language values, making it possible to input language values other than the set language values. .

Further, according to the present invention, the linguistic value static information amount is calculated and the clarity is calculated based on this, so that the clarity (reliability) of the inference result for the linguistic value input can be known.

Description of Embodiments (1) Overall configuration of approximate inference device FIG. 1 shows an example of the overall configuration of the approximate inference device. The approximate inference device includes a knowledge storage device 11. Knowledge synthesis device 1
21 Post-synthesis knowledge storage device 13. Linguistic value input device 14. Fitness calculation device 15. Dynamic information amount calculation device 16. Possibility calculation device 17. Possibility display device 18. Static information amount calculation device 19. Clarity calculation device 20. Clarity storage 21. Clarity adder 22. Clarity display device 23, set language value storage device 31. Membership function creation device 32. and a language value input device 33.

These devices will be explained in detail below.

(2) Knowledge Storage Device The knowledge storage device 11 is a device that stores knowledge input by experts and the like in a format that shows the relationship between events and conclusions. This device can store the knowledge of multiple experts.

Two experts exl stored in the knowledge storage device 11
, ex2 knowledge example is shown below in the form of a rule.

Expert ext: if 20≦r1≦60.0≦f2≦40. the
n cl-11) if' 40≦f1≦80.60
≦12≦100. then c2・=(2) Expert e
x2: H30≦0≦50. 10≦f2≦30. then
cL-(3) if 50≦r1≦70.70≦f
2≦90. then c2=l:4) fL f'2
are events, and these may be called event 1 and event 2, respectively. C1 and c2 are conclusions, and these are sometimes called conclusions 1 and 2, respectively.

Further, a and b expressed as a≦fl≦b are called minimum value and maximum value, respectively.

The above rules are expressed in the form of a table for each expert as follows.

Table 1 Table 2 (3) Knowledge synthesis device The knowledge synthesis device 12 is a device that synthesizes the knowledge of a plurality of experts stored in the knowledge storage device 11 and combines it into one piece of knowledge.

There are various ways to synthesize knowledge, but here we calculate the average value and standard deviation of multiple experts for the maximum and minimum values of each event involved in each conclusion.

The knowledge synthesis process will be described below, taking as an example the knowledge of the two experts described above that leads to the conclusion C1 from the event r1.

From the above rules (Equations (1) and (3)), event 1
The rule for obtaining conclusion 1 (cl) from (fl) is extracted as follows.

Expert ext: 1f 20≦f1≦80 th
en el=15) Expert ex2: if 30
≦f1≦50 then cl=18) The average value m of the minimum values and the average value in '' laX of the maximum values are calculated.

”　1aX 60+50 ...(8) Standard deviation of the minimum value σ and standard deviation of the maximum value ■1n The deviation σ is calculated.

■ax The above-mentioned rules (
For equations (1) to (4), if all the minimum and maximum values of each event involved in each conclusion are examined, a table as shown below is obtained.

Table 3 Generally, in approximate reasoning, a membership function is given to an event. Here, as an example, the special F'! synthesized as described above is used. We will explain how to find the ship function using a Gaussian distribution using basic knowledge.

Using the average value of the minimum value m, the average value of the maximum value mwax'in, the standard deviation of the minimum value σ, and the standard deviation of the maximum value in σ□, the Men/Kuichi-Tsubu function Ct is expressed by the following equation 1. .

...(11) however.

X: Value of input data to the event Φ(X) Degree to which the two-person data X fits the event (degree of fit) Gauss(x): Value of the Gaussian distribution in the input X.

Figure 2 shows an example of a Gaussian distribution. In this case, only the left half of the distribution is used to create the membership function. The position of X at Φ(x)-0,5 is m
or m determined by ■tn
wax and the slope is determined by σ or σ in
Wax Ru.

As an example, the membership function for determining the conclusion C1 from the event fl is created as shown in FIGS. 3a to 3C using values calculated by equations (7) to (lO). In this case, equation (11) becomes as follows.

Figure 3b shows the second term on the right side of equation (11) or (12), and Figure 3C shows the result of subtracting the second term from the first term, that is, equation (11) or equation (12). It represents the membership function expressed by equation (12).

Examples of membership functions for determining conclusions cl and c2 for each event fl and f2 created based on the synthesized knowledge shown in Table 3 are shown in Figures 4a and 4b.
As shown in the figure.

(4) Post-synthesis knowledge storage device The post-synthesis knowledge storage device 13 stores the average value and standard deviation calculated by the knowledge synthesis device 12 in the format shown in Table 3. Since it is not necessary to synthesize knowledge every time an inference is made, the results calculated in advance in this way are memorized. By reading out and using the values in the storage device 13 each time an inference is made, it is possible to speed up the inference process.

(5) Set language value storage device The language value storage device 31 is a device that stores knowledge regarding a plurality of language values for each event input by an expert in the form of a table representing the range of values representing each language value. be. This device 31 is capable of storing the knowledge of a plurality of experts regarding a plurality of language values in each event. Moreover, this language value storage device 31 is.

Since it has the same function as the knowledge storage device 11 described above, these devices 11.31 can also be used for the same purpose.

An example of the knowledge of two experts regarding the language values LL, L2 in event r1 is shown below. Examples of the language values LL and L2 include light rotation of the steering wheel and heavy rotation of the steering wheel.

Expert Exl: fl220≦L1≦30...
(13) 60≦L2≦70
...(14) Expert Ex2: f2: 30≦Ll≦40 ...
(I5) 70≦L2≦80
...(1B) The knowledge of experts regarding the above language values LL and L2 is summarized in the form of a table as follows.

(8) Membership function creation device The membership function creation device 32 synthesizes the knowledge of a plurality of experts regarding the same linguistic value of the same event stored in the setting language value storage device 31, and based on the synthesis result. This creates a membership function.

There are various ways to synthesize linguistic value data.

Here, in the same way as the knowledge described above, the average value and standard deviation of multiple experts are calculated for the minimum and maximum values of each language value.

The following table is obtained by combining the knowledge of the two experts shown in Table 4 and Table 5 for the language values LL and L2, respectively.

Table 6 Linguistic values are expressed as membership functions. Here, in the same way as the membership function given to the above-mentioned event, a membership function of language values is created using a Gaussian distribution. The membership functions created for language values L1 and L2 are shown in FIG.

In this way, the membership function synthesized using the language values of experts etc. set in advance in the set language value storage device 31 is stored in the membership function creation device 32 using data as shown in Table 6. be done.

(7) Language value input device The language value input device 14 is for inputting a desired language value as its representative value. Although various values can be considered as the representative value, here, the midpoint between the average value of the minimum value and the average value of the maximum value of the synthesized linguistic values is taken as the representative value. For example, the 6th
The representative values for language values LL and L2 shown in the table are respectively.

(25+35) / 2-30
...(17) (85+75) / 2-7o
...(18).

As the representative value, it is possible to input not only the representative value of the language values set in advance as described above, but also any arbitrary value. The verbal value input device 14 is a keyboard for humans to enter representative values numerically, a variable resistor (slide volume) as described below for humans to enter representative values by position, or online input of representative values. It consists of a communication device etc. for receiving φ data.

The linguistic value input device 14 outputs the input representative value to the linguistic value interpolation device 33.

The language value interpolation device 33 uses the representative value of the language values given from the language value input device 14 to perform language value interpolation processing as described later, and provides the generated language value data to the language value input device 14. . The linguistic value input device 14 receives the linguistic value data generated by the linguistic value interpolation device 33, sends the linguistic value data as is to the fitness calculation device 15, and also inputs information as to whether a linguistic value belonging to each event has been input. (eg, a value of 1 if input, a value of O if not input) to the clarity adder 22.

When the linguistic value given from the linguistic value input device 14 has already been synthesized by the membership function creation device 32, the interpolation operation in the linguistic value interpolation device 33 is of course unnecessary and the language value is stored in the 2 creation device 32. It is sufficient to read out the language value data as is.

FIG. 6 shows an example of the external appearance of the linguistic value input device 14 configured with a sliding volume. volume 34
, a slide groove and a knob 35 that moves along the slide groove are shown. Further, an input execution button 3B is arranged on the side of the slide groove.

The preset language value Ll shown in FIG.

Assume that L2 expresses the sensations of "light" and "heavy," respectively. The representative values of these language values Ll and L2 are plotted on the sliding volume scale (0, 30° 70,
I (numerical values such as 10 do not necessarily need to be displayed) are made to correspond to predetermined positions, and linguistic values such as "light" and "heavy" are displayed at these predetermined positions.

A person looks at such a display and inputs his or her current feeling by the position of the knob 35. For example, the position of the knob 35 shown in FIG. 6 indicates that the verbal value of ``a little heavy'' is manually set. The position to the left of the display ``light'' is the position to input linguistic values such as very light or extremely light, and the position to the right of the display ``heavy'' is the position to input linguistic values such as very heavy or extremely heavy. This is the position where you enter the value. When a person adjusts the position of the knob 35 and presses the input execution button 36, the output voltage of the slide volume 34 corresponding to the position of the knob 35 at that time is taken in, and that voltage is set to the language value (its representative value). ) is converted to In this way, it becomes possible to input linguistic values that appropriately express the senses of a single person. This linguistic value input device 14 functions as a suitable human-machine interface.

(8) Linguistic value interpolation device The linguistic value interpolation device 33 uses a linguistic value L set in advance.
Using L and L2, linguistic value data represented by the representative value input from the linguistic value input device 14 is generated by interpolation. A linear interpolation method is used here for simplicity. The language value data to be generated are the average value of the minimum value and its standard deviation, and the average value and its standard deviation of the maximum value of the language value. These data will be called parameters.

The representative values of the preset language values LL and L2 are respectively Di and D2. Let the parameters be Pi and P2. Let the representative value of the input language value be old, and the parameter be Pl. Referring to FIG. 7, according to the linear interpolation method.

The input language value parameter Pi is given by the following equation.

Pi= [(P2-PL)/(D2-Di)]
xDi+P2- [(P2-PL) / (D2
-Di) ko XD2... (1
9) - [(P2-PL) / (D2-Di) ) X
Di+pl −[(P2-Pi) / (D2-Di)
] XDI - (20) Now suppose that 60 is given as the representative value D1.

Data representing the language value can be obtained from the above equation (19) or equation (20) as follows.

Average value of the minimum value: 55 Standard deviation of the minimum value: 5 Average value of the maximum value 2 65 Standard deviation of the maximum value; 5 The membership function of the linguistic value represented by such data is shown by the broken line L1 in Figure 5. Become something.

(9) Compatibility calculation device The suitability calculation device 15 calculates the degree of suitability of input language values inputted from the language value data ft14 to each membership function (or conclusion) in a corresponding event. Since the input language value is expressed in the form of a membership function as described above, the goodness of fit is calculated by the old N-MAX operation between the input membership function and each membership function of the corresponding event.

FIG. 8a shows each conclusion e1. of event 1 when language value L1 is given. It shows how the goodness of fit for the membership function that leads to c2 is calculated (the shaded area shows the MIN calculation result between the linguistic value Li and 01).

Let the obtained fitness be m1□1m1°, respectively.

Figure 8b shows the goodness of fit for the membership function that leads to each of the conclusions cl and e2 for event f2 when a certain linguistic value for event 2 (membership function indicated by a broken line) is given. m21. Set to ``22.'' When a membership function representing the input data is given, the goodness of fit is found in the same way.

(10) Dynamic information calculation device and static information calculation device Consider the concept of fuzzy entropy.

When the input language value L (xi) is given, the fuzzy
Entropy Er1 is defined as follows.

This fuzzy entropy is a type of index of information discrimination ability, and when a linguistic value L (xi) is given, the value is small enough that the conclusion can be clearly identified, and the value is large enough that the conclusion can only be vaguely identified. value. In other words, the larger the difference between the degree of conformity m1□ of the input language value L (xi) to the conclusion cl and the degree of conformity "12" of the input language value L (xi) to the conclusion C2, the smaller the value becomes.

The smaller the difference, the larger the value.

In the same way, the fuzzy entropy Ef2 when the input linguistic value L (x2) is given is given by the following formula: Uni and ml, the fitness of the bilingual value L (xi) to the conclusion c1 m1□: Linguistic value The goodness of fit of L (xi) to the conclusion cl (replace Ll with L (xi) in Figure 8a)
21 Degree of suitability m22 of bilingual value L (xi) to conclusion C1: Degree of suitability of linguistic value L (x2) to conclusion C2 The range of possible values of the fuzzy entropy Ef is as shown below.

0≦Ef≦log(n) n: Number of conclusions on event In this example, the number of conclusions on event 1 (fl) is 2 (cl
.

c2), the maximum value of the fuzzy entropy Ef is log(2).

Next, using this fuzzy entropy Efl, we calculate the amount of dynamic information 1 when the input language value L (xi) is given.
Find f'1p(xi). Here, the amount of dynamic information Ir1
D(Xi) is the ability to identify events for evaluating the possibility that a conclusion holds when making inferences, and is the degree of fit m1□ of the input linguistic value L (xl) to the conclusion c1 and the input linguistic value L (xi )'s fitness m12 for the conclusion c2
The larger the difference, the larger the value, and the smaller the difference, the smaller the value.

Therefore, the amount of dynamic information 1flD(xi
) from the maximum fuzzy entropy to the input language value L
Fuzzy entropy Ef when (xi) is given
' Define it as minus l.

Similarly, for event f2, input linguistic value L (x2)
The dynamic information amount calculation unit 16 uses the fitness obtained by the fitness calculation unit 15 to calculate the dynamic information amount when given as follows. The amount of dynamic information is calculated for each event according to the formula.

The dynamic information amount is the input language value L (xi) as described above.

Depends on L (x2). On the other hand, the amount of static information basically does not depend on the input linguistic value, and is calculated from the maximum fuzzy entropy to the fuzzy entropy of all linguistic values that have representative values within the range of the event. The static information amount of the entire event is obtained by subtracting the average of . For example, the static information amount for event 1 is given by the following equation.

(Hereinafter, blank space) 141s=log(2) Compatibility m22(x) for the lesson C1: Linguistic value L(x
)'s suitability for conclusion C2... (25) Similarly, the static information amount for event 2 is calculated by the following formula % Language value L for event fl (x) Compatibility for theory cl Language for event f1 Value L (x) In the conclusion of the linguistic value L (x) for the fitness event r2 for theory C2, as shown in Fig. 9, the representative value X of the linguistic value is changed at 6 intervals, and each linguistic value L An operation to calculate the fuzzy entropy for (x) and find their average (where 0 and δ≦100) As can be seen from equations (25) and (2B), the relationship between the membership functions of events is The greater the overlap, the greater the overlap.

The amount of static information of an event is small, and the smaller the overlap between the membership functions of an event, the larger the amount of static information of an event.

m12(x): m21(x): The conclusion of Small.As shown in Fig. 10b, the language value L (
x) When the spread of The smaller the spread, the larger the static information amount of the event.

The amount of static information indicates the ability of an event's membership function to identify a conclusion, taking into account the spread (ambiguity) of linguistic values. Therefore, even if the knowledge is the same, the amount of static information will decrease if the spread (ambiguity) of linguistic values is large.

The static information amount calculating device 19 is a device that calculates the static information amount of each event with respect to the language value. As mentioned above, the amount of static information for each event represents the ability to identify the event in order to estimate the probability that a conclusion will hold when making inferences by giving an input linguistic value to a certain event. . Therefore, the amount of static information exists equal to the number of events.

The static information amount calculation device 19 changes the representative value of the linguistic value at intervals of δ within the range width of the event for each event, and takes the linguistic value data of each representative value created by the linguistic value interpolation device 33. Then, calculate the degree of fitness for the membership function obtained by the knowledge synthesized once for each linguistic value data, and calculate the amount of static information according to the above-mentioned equations (25) and (28). Remember. Since the static information amount does not depend on input data, it only needs to be calculated once.

(11) Possibility calculation device For each conclusion, calculates the information amount of events such that the sum of the information amount of events related to that conclusion is 1, and the relative strength of the information amount of those events does not change. . This calculated amount of information is called a weight.

For example, if the above-mentioned dynamic information amount is used, each weight will be as follows.

Weight of event 1 relative to conclusion 1: Wf11-Iff, (x
i)/[IffD(xi)+IF5. (x2)]
-(27) Weight of event 2 relative to conclusion 1: wf12I
f2D(x2)/[Ifld(xi)+If2D(
x2) co-(2g) Weight of event 1 on conclusion 2: W f21″″1flp(xi)/[IflD(
xi) + If2o(x2)] ...(29) Weight of event 2 for conclusion 2: wf22-If2D(X2
)/[IfID(Xi)+If2D(X2)co-(ao) Next, calculate the product of these weights and fitness.

The sum of these results for each conclusion is calculated as the probability of the conclusion.

For example, in the above example, the possibility of conclusion 1 −wf Xm +wf Xm
-(31) Possibility of conclusion 2 -wf Xm +w
f Xm −(32).

The possibility calculation device 17 performs the above-mentioned calculation to calculate the probability for each conclusion.

(12) Possibility Display Device This possibility table fl18 manufactured by Hyogo displays the possibilities calculated by the possibility calculation device 17 for each conclusion. The possibility may be displayed for all conclusions, or one or more highly probable conclusions may be displayed. Also, the possibility may be transmitted to another device through one communication, or the possibility may be stored in a memory or a file.

(13) Clarity calculation device The clarity calculation device 20 is a device that calculates the clarity of each event for each conclusion. Here, the clarity of each event for each conclusion indicates the relative discrimination ability of each event when evaluating the possibility of a certain conclusion being established. Therefore, this clarity makes it possible to compare the discriminative ability of multiple events in order to establish a certain conclusion, and to understand which event has a high discriminative ability (contains a large amount of information). Ru. The method for calculating clarity is described below.

First, Table 7 shows the relationship between conclusions, events, and the amount of static information.

Table 7 As can be seen from Table 7, 1; also depending on the amount of static information.

It is possible to compare the discriminative ability of multiple events to establish each conclusion. However, as it is difficult to intuitively understand the relative discrimination ability as it is, we normalize the amount of static information for each conclusion as shown in the table below, and use the normalized value to clearly distinguish each event for each conclusion. Sex CI.

Table 8 however.

C11□-CD12- lff5/ (Iff8+ I
f2s) CD2. - Cl22-1f2s/ (Iff
s+If2s).

In this way, in the clarity calculation device 2o.

The clarity of each event is calculated for each conclusion.

(14) Clarity Storage Device The clarity storage device 21 is a device that stores the clarity of each event for each conclusion calculated by the clarity calculation device 2o.

Clarity calculations do not need to be performed every time an inference is made. Therefore, the clarity calculated when knowledge is synthesized is stored in the clarity storage device 21.

The value stored in the clarity storage device 21 is read out every time an inference is made. This makes it possible to speed up the inference processing.

(15) Clarity Addition Device The clarity addition device 22 is a device that calculates the clarity of an event in which a linguistic value is actually input. Here, for the inference that is actually made, the linguistic value is taken as the summation of the intelligibility of the input events. The sum of this clarity indicates the clarity of the inference result. It can be said that the higher the clarity, the greater the amount of information for deriving the inference result. Therefore, clarity can be used as an indicator to judge the reliability of the inference result itself.

The clarity of the inference result is calculated as follows.

a) When linguistic values are input only for event 1 (fl) - Clarity C111 for the inference result of conclusion 1 (cl)
-CI11□ - Clarity of the inference result of conclusion 2 (c2) cg2-
C11□ b) When linguistic values are input only for event 2 (f2) - Clarity CU for the inference result of conclusion 1 (cl), -
”21 ・Clarity Cj22 for the inference result of Inference 2 (c2)
-”22 C) When linguistic values are input for both Event 1 (fl) and Event 2 (f2) - Clarity cI11 for the inference result of Conclusion 1 (cl)
-CI! +Cl12□-1,0 ・Clarity CN2- for the inference result of conclusion 2 (c2)
CI+ CI, , -1,0 Clarity of inference result Cj! What is the possible range?

0.0≦CD≦1.0. In other words, if one makes an inference by inputting linguistic values for all of the phenomena that can be used to draw a certain conclusion within the knowledge given before making an inference,
The clarity of the conclusion will be 1.0. Also, only some of the events that can be used to draw a certain conclusion.

If a linguistic value is entered, the clarity will be a value between 0.0 and 1.0. At this time, it can be said that if a large number of events with high clarity are used among the events that can be used, the clarity of the conclusion will also be increased, and a highly reliable inference result can be obtained.

(IB) Clarity Display Device The clarity display device 23 is a device that displays the clarity of the inference result (-the possibility mentioned above as an example) calculated by the clarity addition device 22. The clarity may be displayed along with the inference results, or the clarity may be transmitted to another device or stored in memory or a file.

This clarity display displays all conclusions of the inference results. Therefore, if there are multiple conclusions, the clarity corresponding to each conclusion is displayed.

In this way, each time a language value is entered.

By calculating the information amount of the event to which the input language value belongs and displaying the clarity of the inference result, the user can judge the reliability of the inference result.

It goes without saying that each of the devices 11 to 23 and 81 to 33 described above can be realized by a computer including a memory and a display device. For example, knowledge synthesis device 12°, various calculation devices 15.1B, 17.19.20.22.32. H
is preferably realized by a CPU that operates according to a program.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of the overall configuration of an approximate inference device. FIG. 2 is a graph showing a Gaussian distribution. Figures 3a to 3c are graphs showing how membership functions are formed. Figures 4a and 4b are graphs showing the membership functions obtained for each event. FIG. 5 is a graph representing membership functions of linguistic values created by interpolation processing. FIG. 6 is an external view showing an example of a linguistic value input device. FIG. 7 is a graph showing the concept of linear interpolation. FIGS. 8a and 8b are graphs showing how the goodness of fit is determined. FIG. 9 is a graph showing how the linguistic value static information amount is calculated. FIGS. 10a and 10b are graphs showing how the amount of static information changes depending on the spread of language values. 11...Knowledge storage device. 12...Knowledge synthesis device. 13... Post-synthesis knowledge storage device. 14... Linguistic value input device. 15...Fitness calculation device. 1B...Dynamic information calculation device. 17... Possibility calculation device. 18... Possibility display device. I9: Static information calculation device. 20... Clarity calculation device. 21... Clarity memory. 22... Clarity addition device. 23... Clarity display device. 31... Setting language value storage device. 32...Membership function creation device. 33... Linguistic value interpolation device. Below

Claims (1)

[Scope of Claims] (1) For each of a plurality of set language values related to an event, each set language value is handled by statistically processing data given by multiple experts indicating the meaning range of each set language value. means for creating and storing data representing a membership function for the set language value; An approximate inference device comprising linguistic value interpolation means that performs interpolation processing to create data representing membership functions of input linguistic values. (2) A conclusion can be drawn by applying data representing a membership function created by interpolation processing on data for specifying input linguistic values to preset expert knowledge representing the relationship between an event and a conclusion. The approximate inference device according to claim 1, further comprising approximate inference means for inferring the possibility of . (3) Data representing the membership function created by the above linguistic value interpolation means for each specific data by changing the data for specifying the linguistic value within the range of the event at regular intervals, and the relationship between the event and the conclusion. Claim (1) further comprising static information amount calculation means for calculating the amount of linguistic value static information for each event using expert knowledge representing the
The approximate inference device described in . (4) A clarity calculation means that calculates the clarity of each event for each conclusion using the calculated amount of static information, and a language value that is actually input using the clarity obtained from the clarity calculation means. The approximate inference device according to claim 3, further comprising: clarity addition means for calculating the clarity of the event. (5) By statistically processing data provided by multiple experts indicating the range of meaning for each of multiple language settings values related to an event, the membership function corresponding to each language value is expressed. a means for creating and storing data; a language value input means for inputting data specifying a language value to be input; An approximate inference device comprising linguistic value interpolation means for performing interpolation processing on the specific data using data representing a membership function to create data representing a membership function of an input linguistic value. (8) By applying the data representing the membership function created by the above interpolation process regarding the input linguistic values to the preset expert knowledge representing the relationship between the event and the conclusion, the possibility of the conclusion is determined. The approximate inference device according to claim (5), comprising: approximate inference means for inferring . (7) By statistically processing data provided by multiple experts indicating the range of meaning for each of multiple language settings values related to an event, the membership function corresponding to each language value is expressed. A means of creating and storing data. When data for specifying a language value is given, interpolation processing is performed on the specified data using the data representing the membership function of the set language value to create data representing the membership function of the input language value. a linguistic value interpolation means for specifying a linguistic value within the range of an event, and a member created by the linguistic value interpolation means with respect to each specific data by changing data for specifying a linguistic value within the range of an event at regular intervals and feeding it to the linguistic value interpolation means. An approximate inference device comprising static information amount calculation means for calculating a linguistic value static information amount for each event using data representing a ship function and expert knowledge representing a relationship between an event and a conclusion. (8) A clarity calculation means that calculates the clarity of each event for each conclusion using the calculated amount of static information, and a language value that is actually input using the clarity obtained from the clarity calculation means. The approximate inference device according to claim 7, further comprising: clarity addition means for calculating the clarity of the event.