Fuzzy-boolean multi-stage inference apparatus

• US 5,191,638 A
• Filed: 03/29/1990
• Issued: 03/02/1993
• Est. Priority Date: 03/31/1989
• Status: Expired due to Term

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Naito, Eiichi, Hayashi, Isao, Nomura, Hiroyoshi, Wakami, Noboru
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/501,037
• Time in Patent Office: 1,069 Days
• Field of Search: 364/513, 364/207, 395/61, 395/900, 395/11, 395/51, 395/54
• US Class Current: 706/10
• CPC Class Codes: G06N 5/048 Fuzzy inferencing
• CPC Class Codes: G06N 7/04 Physical realisation
• CPC Class Codes: Y10S 706/90 Fuzzy logic

Abstract

An inference conducted by use of fuzzy knowledge can be directly written in inference rules. It is possible to use ambiguous language expressions and freely describe flexible knowledge, although ambiguous, in a knowledge base. The user specifies execution priority orders or conviction degrees of inference rules such that even when an antecedent contains a confirmative or determinant action, a contention between rules is prevented. An intermediate results of a fuzzy inference is passed in a form of a real number or a fuzzy label to the next inference to achieve a fuzzy multi-stage inference. A multi-stage inference conducting a backtrack operation based on fuzzy knowledge is implemented to reflect a user'"'"'s ambiguous knowledge. An inference method associated with a best value retrieval function is provided, in contrast to the conventional method in which a fuzzy inference and a binary logic inference are continuously achieved from an upper-level retrieval or search point to a lower-level retrieval point in a one-directional manner.

Claims

1. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules;

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory; and

an inference rule selector for determining, by using an evaluation formula, a priority inference rule to be executed by said inference means in a succeeding step in said multi-step inference processing;

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory;

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules and Boolean logic inference rules,

said inference means outputs said fuzzy inference rules and said Boolean logic inference rules in one of said knowledge sources to said working region for use in execution of fuzzy logic and Boolean logic inferences, and

said inference rule selector comprises a priority memory in which said priority of inference rules is stored;

a compatibility processor for deriving a compatibility of said inference rule to said IF parts; and

an inference rule determination means for deriving an inference rule having a priority of more than a predetermined value and a highest priority in said priority memory.

2. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules;

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory; and

an inference rule selector for determining, by using an evaluation formula, a priority inference rule to be executed by said inference means in a succeeding step in said multi-step inference processing;

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory;

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules and Boolean logic inference rules,

said inference means outputs said fuzzy inference rules and said Boolean logic inference rules in one of said knowledge sources to said working region for use in execution of fuzzy logic and Boolean logic inferences, and

said inference rule selector comprises a reliability memory in which a representation of reliability of said inference rule is stored;

a compatibility processor for determining the compatibility of said IF parts;

a priority processor to determine priority between said inference rule in said compatibility processor and said reliability in said reliability memory to derive a priority inference rule; and

an execution rule determination means for deriving a fuzzy inference rule having a highest priority in said priority processor.

3. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory;

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules or Boolean logic inference rules, and

said inference means outputs fuzzy inference rules of some of said knowledge sources to said working region for use in executing said fuzzy logic and/or Boolean logic inferences,

said apparatus further comprising an intermediate result determination means for determining an intermediate result and a final result to derive a priority fuzzy inference rule by using a fuzzy inference to determine an inference rule to be executed next by said inference means.

4. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory,

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules or Boolean logic inference rules, and

said inference means outputs fuzzy inference rules of some of said knowledge sources to said working region for use in executing said fuzzy logic and/or Boolean logic inferences,

said apparatus further comprising an inference re-execution starting means for determining timing to start a re-execution of said multi-stage inference and an inference rule to be re-executed by using knowledge comprising fuzzy sets.

5. A fuzzy multi-stage inference to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory,

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules or Boolean logic inference rules, and

said inference means outputs fuzzy inference rules of some of said knowledge sources to said working region for use in executing said fuzzy logic and/or Boolean logic inferences,

said apparatus further comprising an inference concluding means for determining an ending of said multi-stage inference apparatus process by using knowledge described in fuzzy sets.

6. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory,

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules and Boolean logic inference rules, and

said inference means outputs said fuzzy inference rules and said Boolean logic inference rules in one of said knowledge sources to said working region for use in execution of fuzzy logic and Boolean logic inferences,

said apparatus further comprising an intermediate result determination means for determining an intermediate result and a final result to derive a priority fuzzy inference rule by using a fuzzy inference to determine an inference rule to be executed next by said inference means.

7. A fuzzy multi-stage inference apparatus according to claim 6 wherein said intermediate result determination means comprises an intermediate result inference means for executing said fuzzy inference and an intermediate result working region to store results of said fuzzy inference;

and

said intermediate result inference means re-executes inferences after converting an intermediate result of fuzzy inference into real values.

8. A fuzzy multi-stage inference apparatus according to claim 7 wherein said intermediate result determination means determines real values of said intermediate result by deriving a centroid of a fuzzy set determined by said fuzzy inference.

9. A fuzzy multi-stage inference apparatus according to claim 6 wherein said intermediate result determination means comprises an intermediate result inference means for executing said fuzzy inference and an intermediate result working region to store results of said fuzzy inference;

and

said intermediate result inference means re-executes said inferences based upon both real values of said intermediate result of fuzzy inference and newly inputted data.

10. A fuzzy multi-stage inference apparatus according to claim 9 wherein said intermediate result determination means determines real values of said intermediate result by deriving a centroid of a fuzzy set determined by said fuzzy inference.

11. A fuzzy multi-stage inference apparatus according to claim 6 wherein said intermediate result determination means comprises an intermediate result inference means for executing said fuzzy inference and an intermediate result working region to store said inference results;

and

said intermediate result inference means re-inputs a fuzzy set obtained by said fuzzy inference into said intermediate result inference means after said fuzzy set is substituted by fuzzy labels in order to execute multi-stage inferences by using said fuzzy labels.

12. A fuzzy multi-stage inference apparatus according to claim 11 wherein a fuzzy label for minimizing the difference between an intermediate result of a fuzzy set and an area of a fuzzy label intermediately obtained by fuzzy inference is selected.

13. A fuzzy multi-stage inference apparatus according to claim 6 wherein said intermediate result determination means comprises an intermediate result inference means for executing said fuzzy inference and an intermediate results working region to store said-inference results;

and

said intermediate result inference means re-inputs a fuzzy set obtained by said fuzzy inference into said intermediate result inference means after said fuzzy set is substituted by fuzzy labels in order to execute multi-step inferences by considering both said fuzzy labels and newly inputted data.

14. A fuzzy multi-stage inference apparatus according to claim 13 wherein a fuzzy label for minimizing the difference between an intermediate result of a fuzzy set and an area of a fuzzy label intermediately obtained by fuzzy inference is selected.

15. A fuzzy multi-stage inference apparatus according to claim 6 wherein said intermediate result determination means comprises an intermediate result inference means for executing said fuzzy inference and an intermediate working region to store said inference results;

said intermediate result inference means comprises a multi-step inference means for converting said intermediate results of said fuzzy inference into real values; and

a multi-step inference means for substituting said fuzzy sets obtained by said fuzzy inference by fuzzy labels.

16. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory,

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules or Boolean logic inference rules, and

said inference means outputs fuzzy inference rules of some of said knowledge sources to said working region for use in executing said fuzzy logic and/or Boolean logic inferences,

said apparatus further comprising an inference re-execution starting means for determining timing to start a re-execution of said multi-stage inference and an inference rule to be re-executed by using knowledge comprising fuzzy sets.

17. A fuzzy multi-stage inference apparatus according to claim 16 wherein said inference re-execution starting means commences said re-execution of inference only when a compatibility of said input data to said IF parts of fuzzy inference rules satisfies a predetermined criterion.

18. A fuzzy multi-stage inference apparatus according to claim 17 further comprising an inference re-execution rule searching means for determining inference rules to be searched at said re-execution of inference.

19. A fuzzy multi-stage inference apparatus according to claim 18 wherein said inference re-execution rule searching means searches for inference rules in an order of higher adaptability of a given number of inference rules comprising an antecedent.

20. A fuzzy multi-stage inference apparatus according to claim 18 wherein said inference re-execution rule searching means searches for an inference rule having a next highest adaptability to that of a final inference rule among inference rules in aid knowledge source and/or a preceding knowledge source.

21. A fuzzy multi-stage inference apparatus according to claim 18 wherein said inference re-execution rule searching means comprises a fuzzy result evaluation means for determining user satisfaction of said inference result attained by using knowledge described in a fuzzy set;

and

a best solution searching processor for finding a best command from THEN parts of said inference rules which led to executed inference rules.

22. A fuzzy multi-stage inference apparatus according to claim 18 wherein said inference re-execution rule searching means comprises a fuzzy result evaluation means for determining user satisfaction with said inference results attained by using knowledge described in a fuzzy set;

and

a maximum evaluation value searching processor by which a rule with a highest evaluation value is searched as a result of fuzzy inference among a knowledge source to which said inference rule which led to an executed inference rule belongs.

23. A fuzzy multi-stage inference apparatus according to claim 18 wherein said inference re-execution-rule searching means comprises a fuzzy result evaluation means for evaluating user satisfaction of said inference results attained by using a knowledge described in a fuzzy set;

an exponential processor which conducts exponential processing by using said adaptability of said antecedent and an evaluation value derived by said fuzzy result evaluation part; and

a maximum processing result processor for searching for an inference rule by which a highest processing result can be derived among knowledge that an inference rule which led to an executed inference rule belongs.

24. A fuzzy multi-stage inference apparatus according to claim 23, wherein said evaluation means is executed by a multiplication process executed by an exponential processor.

25. A fuzzy multi-stage inference apparatus according to claim 23, wherein said evaluation is executed by a minimum deriving process executed by an exponential processor.

26. A fuzzy multi-stage inference apparatus according to claim 16 wherein said inference re-execution starting means comprises a fuzzy result evaluation means for evaluating a user satisfaction of an inference result;

and

a back-track determination means for determining a re-execution of said inference determines when an evaluation derived by said fuzzy result evaluation part has satisfied a predetermined criterion.

27. A fuzzy multi-stage inference apparatus

according to claim 26 further comprising an inference re-execution rule searching means for determining inference rules to be searched at said re-execution of inference.

28. A fuzzy multi-stage inference apparatus to be employed in information processing systems for diagnosis or recognition purposes, comprising:

an inference rule memory for storing fuzzy inference rules comprising IF parts, THEN parts, and Boolean logic inference rules; and

an inference means for deriving control commands for executing multi-step inference processing comprising one of (i) fuzzy logic inferences and (ii) fuzzy logic and Boolean logic inferences in series or in parallel by utilizing said rules stored in said inference rule memory,

wherein;

said inference means outputs a portion of the fuzzy inference rules stored in said inference rule memory to a working region of said memory,

said inference means executes said multi-step inference processing by using only said inference rules outputted to said working region,

said inference rules are divided into plural knowledge sources each of which comprises fuzzy inference rules and Boolean logic inference rules, and

said inference means outputs said fuzzy inference rules and said Boolean logic inference rules in one of said knowledge sources to said working region for use in execution of fuzzy logic and Boolean logic inferences,

said apparatus further comprising an inference concluding means for determining an ending of said multi-stage inference apparatus process by using knowledge described in fuzzy sets.

29. A fuzzy multi-stage inference apparatus according to claim 28 wherein said inference concluding means comprises a fuzzy result evaluation means for determining user satisfaction of a result of said multi-stage inference;

and

a conclusion determining means for determining an ending of said multi-stage inference when an evaluation of said inference has satisfied a predetermined criterion.