US5283838A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Neural network apparatus
Assignee: Sharp Corp
•
Filing: 1991-04-15
•
Granted: 1994-02-01
•
Expires: 2011-04-15
Abstract
Abstract When performing learning for a neural network, a plurality of learning vectors which belong to an arbitrary category are used, and self-organization learning in the category is carried out. As a result, the plurality of learning vectors which belong to the category are automatically clustered, and the contents of weight vectors in the neural network are set to representative vectors which exhibit common features of the learning vectors of each cluster. Then, teacher-supervised learning is carried out for the neural network, using the thus set contents of the weight vectors as initial values thereof. In the learning process, an initial value of each weight vector is set to the representative vector of each cluster obtained by clustering. Therefore, the number of calculations required until the teacher-supervised learning is converged is greatly reduced.
Inventors
- 👤 Fumio Togawa
- 👤 Toru Ueda
- 👤 Takashi Aramaki
- 👤 Yasushi Ishizuka
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-25 00:43
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/10; G06F; G06F18/00; G06F18/20; G06F18/24; G06F18/241; G06F18/2413; G06F18/24133; G06F18/24137
Claims Analysis
Total Claims: 13
Claim 1
Independent Claim
127 words
A neural network apparatus comprising self-organization learning means for performing self-organization learning to a plurality of learning vectors in a category to which said learning vectors belong, a plurality of weight vectors defined in said neural network apparatus, said weight vectors being allocated to said category, said self-organization learning means having modification means for modifying contents of a predetermined number of said weight vectors which are in a vicinity of one of said learning vectors, toward said one learning vector, said apparatus further comprising an initial contents setting means for setting said modified contents of predetermined number of said weight vectors as initial contents of said predetermined number of said weight vectors, teacher-supervised learning means for performing teacher-supervised learning to said learning vectors, using said initial contents.
Claim 2
Dependent Claim
127 words
A neural network apparatus comprising self-organization learning means for performing self-organization learning to a plurality of learning vectors in a category to which said learning vectors belong, a plurality of weight vectors defined in said neural network apparatus, said weight vectors being allocated to said category, said self-organization learning means having modification means for modifying contents of a predetermined number of said weight vectors which are in a vicinity of one of said learning vectors, toward said one learning vector, said apparatus further comprising an initial contents setting means for setting said modified contents of predetermined number of said weight vectors as initial contents of said predetermined number of said weight vectors, teacher-supervised learning means for performing teacher-supervised learning to said learning vectors, using said initial contents.
Claim 3
Dependent Claim
12 words
An apparatus according to claim 1, wherein said predetermined number is one.
Claim 4
Dependent Claim
14 words
An apparatus according to claim 1, wherein said predetermined number is two or more.
Claim 5
Dependent Claim
25 words
An apparatus according to claim 1, wherein said self-organization learning means performs self-organization learning in every category to which said plurality of learning vectors belong.
Claim 6
Dependent Claim
15 words
An apparatus according to claim 1, wherein said apparatus comprises a Kohonen type neural network.
Claim 7
Dependent Claim
27 words
An apparatus according to claim 5, wherein said apparatus comprises: output nodes allocated to said category; and an input node through which said learning vectors are input.
Claim 8
Dependent Claim
15 words
An apparatus according to claim 1, wherein said apparatus comprises a perceptron type neural network.
Claim 9
Dependent Claim
28 words
An apparatus according to claim 7, wherein said apparatus comprises: intermediate layer nodes allocated to said category; and input layer nodes through which said learning vectors are input.
Claim 10
Independent Claim
112 words
A method of performing learning in a neural network apparatus, comprising the steps of: performing self-organization learning to a plurality of learning vectors in a category to which said learning vectors belong, a plurality of weight vectors defined in said neural network apparatus, said weight vectors being allocated to said category, while modifying contents of a predetermined number of said weight vectors which are in a vicinity of one of said learning vectors, toward said one learning vector; setting said modified contents of predetermined number of said weight vectors as initial contents of said predetermined number of said weight vectors; and performing teacher-supervised learning to said learning vectors, using said initial contents.
Claim 11
Independent Claim
112 words
A method of performing learning in a neural network apparatus, comprising the steps of: performing self-organization learning to a plurality of learning vectors in a category to which said learning vectors belong, a plurality of weight vectors defined in said neural network apparatus, said weight vectors being allocated to said category, while modifying contents of a predetermined number of said weight vectors which are in a vicinity of one of said learning vectors, toward said one learning vector; setting said modified contents of predetermined number of said weight vectors as initial contents of said predetermined number of said weight vectors; and performing teacher-supervised learning to said learning vectors, using said initial contents.
Claim 12
Dependent Claim
19 words
A method according to claim 9, wherein said self-organization learning step and said teacher-supervised learning step are conducted simultaneously.
Claim 13
Dependent Claim
26 words
A method of performing learning in a neural network apparatus according to claim 9, wherein said teacher-supervised learning step are conducted after said self-organization learning step.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1997-01-31 | 1997-08-02 | 1998-02-01 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2001-01-31 | 2001-08-02 | 2002-02-01 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2005-01-31 | 2005-08-02 | 2006-02-01 | $7700.00 | Window Closed / Lapsed |