US5311601A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Hierarchical pattern recognition system with variable selection weights
Assignee: TRUSTEES OF BOSTON UNIVERSITY A Corp OF
•
Filing: 1991-01-11
•
Granted: 1994-05-10
•
Expires: 2011-01-11
Abstract
Abstract In a pattern recognition system, input signals are applied to a short term feature representation field of nodes. A pattern from the short term feature representation field selects at least one category node in a category representation field. The selected category then generates a template pattern. With an insufficient match between the input pattern and template pattern, the category selection is reset. Category selection is based on selection weights which are initially set equal to long term memory weights. After reset, however, selections weights are reduced. Reduction is greatest at those nodes where excitation in F 2 was greater prior to reset. The category representation field is of the same form as the field which receives the input and may itself serve as an input to a higher level pattern recognition system.
Inventors
- 👤 Gail A. Carpenter
- 👤 Stephen Grossberg
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-31 17:41
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/0409; G06F; G06F18/00; G06F18/20; G06F18/23; G06F18/232; G06F18/2321; G06F18/23211
Claims Analysis
Total Claims: 30
Claim 1
Independent Claim
139 words
A pattern recognition data processing system comprising: a short term feature representation field of nodes supporting signals derived from input signals defining an input pattern; a category representation field of a plurality of nodes representing pattern categories; means for selecting at least one category node of the category representation field based on a pattern from the short term feature representation field; and means for resetting a category selection to cause continued search for an acceptable category selection; wherein the means for selecting generates activation signals at nodes of the category representation field which are a function of signals from nodes of the feature representation field weighted by selection weights, the selection weights being initially set for a category selection search by long term memory but being reduced when selecting at least one category after reset of a category selection.
Claim 2
Dependent Claim
139 words
A pattern recognition data processing system comprising: a short term feature representation field of nodes supporting signals derived from input signals defining an input pattern; a category representation field of a plurality of nodes representing pattern categories; means for selecting at least one category node of the category representation field based on a pattern from the short term feature representation field; and means for resetting a category selection to cause continued search for an acceptable category selection; wherein the means for selecting generates activation signals at nodes of the category representation field which are a function of signals from nodes of the feature representation field weighted by selection weights, the selection weights being initially set for a category selection search by long term memory but being reduced when selecting at least one category after reset of a category selection.
Claim 3
Dependent Claim
27 words
A pattern recognition system as claimed in claim 1 wherein said selection weights are reduced as a function of the at least one category which is selected.
Claim 4
Dependent Claim
36 words
A pattern recognition system as claimed in claim 1 wherein, in generating the activation signals, the signals from nodes from the feature representation field are further weighted by corresponding activation signals to contrast enhance category selection.
Claim 5
Dependent Claim
26 words
A pattern recognition system as claimed in claim 1 further comprising means for adapting initial selection weights set by long term memory to the input pattern.
Claim 6
Dependent Claim
31 words
A pattern recognition system as claimed in claim 1 further comprising means for generating a long term memory template, based on the selected category, to be matched against the input pattern.
Claim 7
Dependent Claim
23 words
A pattern recognition system as claimed in claim 1 further comprising means for adapting the long term memory template to the input signals.
Claim 8
Dependent Claim
67 words
A pattern recognition system as claimed in claim 1 wherein the category representation field comprises a plurality of subfields of nodes supporting short term signals and feedback of short term signals within the short term field, each of the short term signals of at least one subfield being based on signals specific to a category and also on nonspecific signals generated from across the short term field.
Claim 9
Dependent Claim
39 words
A pattern recognition system as claimed in claim 1 further comprising a second category representation field and means for selecting at least one category of the second category representation field based on pattern from the first category representation field.
Claim 10
Independent Claim
129 words
A method of pattern recognition in a pattern recognition data processing system comprising: storing input signals defining an input pattern in a short term feature representation field of nodes; selecting at least one category node of a category representation field comprised of a plurality of nodes representing pattern categories based on a pattern from the short term feature representation field; and resetting a category selection; in selecting at least one category node, generating activation signals at nodes of the category representation field which are a function of signals from nodes of the feature representation field weighted by selection weights, the method further comprising initially setting the selection weights by long term memory and reducing the selection weights in selecting at least one category after reset of a category selection.
Claim 11
Independent Claim
129 words
A method of pattern recognition in a pattern recognition data processing system comprising: storing input signals defining an input pattern in a short term feature representation field of nodes; selecting at least one category node of a category representation field comprised of a plurality of nodes representing pattern categories based on a pattern from the short term feature representation field; and resetting a category selection; in selecting at least one category node, generating activation signals at nodes of the category representation field which are a function of signals from nodes of the feature representation field weighted by selection weights, the method further comprising initially setting the selection weights by long term memory and reducing the selection weights in selecting at least one category after reset of a category selection.
Claim 12
Dependent Claim
25 words
A method as claimed in claim 9 wherein the selection weights are reduced as a function of the at least one category which is selected.
Claim 13
Dependent Claim
32 words
A method as claimed in claim 9 further comprising, in generating the activation signals, weighting signals from nodes of the feature representation field by corresponding activation signals to contrast enhance category selection.
Claim 14
Dependent Claim
22 words
A method as claimed in claim 9 further comprising adapted initial selection weights set by long term memory to the input pattern.
Claim 15
Dependent Claim
36 words
A method as claimed in claim 9 further comprising generating a long term memory template, based on the selected category and matching the memory template against the input pattern to determine reset of a category selection.
Claim 16
Dependent Claim
19 words
A method as claimed in claim 9 further comprising adapting the long term memory template to the input signals.
Claim 17
Dependent Claim
66 words
A method as claimed in claim 9 further comprising, in the category representation field, supporting short term signals in a plurality of subfields and providing feedback of short term signals within the short term field, each of the short term signals of at least one subfield being based on signals specific to a category and also on nonspecific signals generated from across the short term field.
Claim 18
Dependent Claim
30 words
A method as claimed in claim 9 further comprising selecting at least one category of a second category representation field based on a pattern from the first category representation field.
Claim 19
Dependent Claim
115 words
A pattern recognition data processing system comprising: a short term feature representation field of nodes supporting signals derived from input signals defining an input pattern and template signals defining a long term memory template; means for selecting at least one category of a category representation field of nodes based on a pattern from the short term feature representation field; means for generating the long term template based on the at least one selected category; and means for adapting category selection and the long term memory template to the input signals; the category representation field comprising a plurality of subfields of nodes supporting short term signals and feedback of short term signals within the category representation.
Claim 20
Dependent Claim
115 words
A pattern recognition data processing system comprising: a short term feature representation field of nodes supporting signals derived from input signals defining an input pattern and template signals defining a long term memory template; means for selecting at least one category of a category representation field of nodes based on a pattern from the short term feature representation field; means for generating the long term template based on the at least one selected category; and means for adapting category selection and the long term memory template to the input signals; the category representation field comprising a plurality of subfields of nodes supporting short term signals and feedback of short term signals within the category representation.
Claim 21
Dependent Claim
77 words
A pattern recognition system as claimed in claim 17 wherein at least one category is selected based on a pattern from the short term feature representation field weighted by selection weights, the system further comprising means for resetting a category selection, the selection weights being initially set by long term memory but being reduced, as a function of at least one category which is selected, in selecting at least one category after reset of a category selection.
Claim 22
Dependent Claim
36 words
A pattern recognition system as claimed in claim 17 wherein, in generating the activation signals, the signal from nodes from the feature representation field are further weighted by corresponding activation signals to contrast enhance category selection.
Claim 23
Dependent Claim
41 words
A category selection system as claimed in claim 17 wherein each of the short term signals of at least one subfield is based on signals specific to a category and also on nonspecific signals generated from across the short term field.
Claim 24
Dependent Claim
40 words
A pattern recognition system as claimed in claim 17 further comprising a second category representation field, and means for selecting at least one category of the second category representation field based on a pattern from the first category representation field.
Claim 25
Independent Claim
111 words
A method of pattern recognition in a pattern recognition data processing system comprising storing input signals in a short term representation field of nodes; selecting at least one category of a category representation field comprised of a plurality of nodes representing pattern categories based on a pattern from the short term feature representation field; generating a long term template based on the at least one selected category; and adapting category selection and the long term memory template to the input signals; the method further comprising, in the category representation field, generating short term signals in a plurality of subfields and providing feedback of short term signals within the category representation field.
Claim 26
Independent Claim
111 words
A method of pattern recognition in a pattern recognition data processing system comprising storing input signals in a short term representation field of nodes; selecting at least one category of a category representation field comprised of a plurality of nodes representing pattern categories based on a pattern from the short term feature representation field; generating a long term template based on the at least one selected category; and adapting category selection and the long term memory template to the input signals; the method further comprising, in the category representation field, generating short term signals in a plurality of subfields and providing feedback of short term signals within the category representation field.
Claim 27
Dependent Claim
94 words
A method as claimed in claim 22 further comprising selecting the at least one category based on a pattern from the short term feature representation field weighted by selection weights, and resetting the category selection with an insufficient match between the long term template and the input pattern in the feature representation field, the method further comprising initially setting the selection weights by long term memory but reducing the selection weights, as a function of at least one category which is selected, in selecting at least one category after reset of a category selection.
Claim 28
Dependent Claim
59 words
A method as claimed in claim 22 further comprising generating activation signals at nodes of the category representation field which are a function of signals from nodes of the feature representation field weighted by selection weights, the method further comprising further weighting signals from nodes from the feature representation field by corresponding activation signals to contrast enhance category selection.
Claim 29
Dependent Claim
39 words
A method as claimed in claim 22 wherein each of the short term signals of at least one subfield is based on signals specific to a category and also on nonspecific signals generated from across the short term field.
Claim 30
Dependent Claim
30 words
A method as claimed in claim 22 further comprising selecting at least one category of a second category representation field based on a pattern from the first category representation field.
References Cited (Backward Citations)
US4914704A
Cited
WO1988010476A1
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1997-05-09 | 1997-11-08 | 1998-05-10 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2001-05-09 | 2001-11-08 | 2002-05-10 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2005-05-09 | 2005-11-08 | 2006-05-10 | $7700.00 | Window Closed / Lapsed |