Knowledge information processing system including two associative memory groups each including a plurality of associated memory units

• US 5,418,886 A
• Filed: 01/05/1993
• Issued: 05/23/1995
• Est. Priority Date: 01/07/1992
• Status: Expired due to Term

• Current Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Inventors: Kyuma, Kazuo, Tai, Shuichi, Oita, Masaya
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Dorvil, Richemond
• Application Number:
  US08/000,624
• Time in Patent Office: 868 Days
• Field of Search: 395/25, 395/26, 395/11, 395/24, 395/23
• US Class Current: 706/25
• CPC Class Codes: G06F 9/4496 Unification in logic programming

Abstract

The knowledge information processing system includes a first and second associative memory unit groups. The associative memory units of the first associative memory unit group 1 associates a plurality of distinct features to the input pattern 20. In response to the combination of associated outputs from the first associative memory unit group 1, the second associative memory unit group 4 evaluates and associates features corresponding to those associated by the first associative memory unit group 1. The logical operation unit group 6 compares the associated outputs of the first associative memory unit group 1 with those of the second to judge whether or not corresponding associated outputs agree with each other. If the corresponding associated outputs do not agree, the logical operation unit group 6 outputs feedback information items for correcting the energy functions of the first associative memory unit group 1, thereby repeating the association and the evaluation process.

Claims

1. A knowledge information processing system comprising:

a first associative memory unit group comprising a plurality of associative memory units according to a neural network model for associating distinct features in response to an input pattern, said associative memory units of said first associative memory unit group generating associated outputs corresponding to said distinct features based on association; and

a second associative memory unit group comprising a plurality of associative memory units according to a neural network model each coupled to outputs of said associative memory units of said first associative memory unit group, said second associative memory unit group evaluating a combination of said associated outputs of said associative memory units of said first associative memory unit group, each of said associative memory units of said second associative memory unit group generating associated outputs which are input to respective associative memory units of said first associative memory unit group for adding a constraint condition to an energy function of the neural network model of the respective associative memory unit of said first associative memory unit group;

wherein said energy function of each associative memory unit of said first associative memory unit group is updated by said associated output of said respective associative memory unit of said second associative memory unit group, such that said associating by said first associative memory unit group and said evaluating by said second associative memory unit group and said updating of energy function of said associative memory units of said first associative memory unit group are repeated with respect to said input pattern.

2. The knowledge information processing system of claim 1, wherein:

the first associative memory unit group comprises a plurality of associative memory units according to a Hopfield neural network model; and

the second associative memory unit group comprises a plurality of associative memory units according to a feed-forward type of neural network model.

3. The knowledge information processing system of claim 1, wherein training of the first associative memory unit group is performed independently of training of the second associative memory unit group.

4. A knowledge information processing system comprising:

a first associative memory unit group comprising a plurality of associative memory units according to a neural network model for associating distinct features in response to an input pattern, said associative memory units of said first associative memory unit group generating associated outputs corresponding to said distinct features based on association;

a second associative memory unit group comprising a plurality of associative memory units according to a neural network model each coupled to outputs of said associative memory units of said first associative memory unit group for evaluating a combination of said associated outputs of said associative memory units of said first associative memory unit group, said associative memory units of said second associative memory unit group generating associated outputs corresponding to a combination of said distinct features; and

a logical operation unit group comprising a plurality of logical operation units coupled to said first associative memory unit group and second associative memory unit group, for comparing said associated outputs of said first associative memory unit group with said associated outputs of said second associative memory unit group, said logical operation units of said logical operation unit group generating feedback information items which are input to respective associative memory units of said first associative memory unit group for updating an energy function of the neural network model of the respective associative memory unit of said first associative memory unit group;

wherein said associating by said first associative memory unit group and said evaluating by said second associative memory unit group and said updating of said energy function of said associative memory units of said first associative memory unit group based on said feedback information items are repeated with respect to said input pattern.

5. The knowledge information processing system of claim 4, wherein:

the first associative memory unit group comprises a plurality of associative memory units according to a Hopfield neural network model; and

the second associative memory unit group comprises a plurality of associative memory units according to a feed-forward type of neural network model.

6. The knowledge information processing system of claim 4, wherein training of the first associative memory unit group is performed independently of training of the second associative memory unit group.

7. The knowledge information processing system of claim 6, wherein:

each of the first and second associative memory unit groups is previously trained; and

the updating of the energy function of the associative memory units of the first associative memory unit group based on the feedback information items is not part of the training of the first associative memory unit group.