Machine learning
First Claim
Patent Images
1. A machine comprising:
- (a) a memory which stores a collection of templates, the collection of templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, and(b) a learning component that constructs said templates by at leasttesting a first set of templates by at leastcategorizing an example of an item having a known categorization into one of the set of categories based on computed relationships between each of a first set of templates and the example,constructing a second set of templates based on the testing and first set of templates,forming the collection of the templates stored in the memory based on performing the testing and the constructing one or more times.
3 Assignments
0 Petitions
Accused Products
Abstract
This invention relates to the machine recognition and learning of predetermined categories and more generally, to the representation of patterns, information and knowledge in computational applications. A method of learning categories is an important component of advanced software technology. This invention has applications in the following areas: bioinformatics, document classification, document similarity, financial data mining, goal-based planners, handwriting and character recognition, information retrieval, natural language processing, natural language understanding, pattern recognition, search engines, strategy based domains such as business, military and games, and vision recognition.
66 Citations
92 Claims
-
1. A machine comprising:
-
(a) a memory which stores a collection of templates, the collection of templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, and (b) a learning component that constructs said templates by at least testing a first set of templates by at least categorizing an example of an item having a known categorization into one of the set of categories based on computed relationships between each of a first set of templates and the example, constructing a second set of templates based on the testing and first set of templates, forming the collection of the templates stored in the memory based on performing the testing and the constructing one or more times. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A machine comprising:
a memory that stores one or more category spaces, a collection of templates, the collection of templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, and a learning component that constructs said templates.
-
18. A machine comprising:
-
(a) a memory which stores a collection of one or more category distinguishing elements, which will be referred to as one or more templates, the one or more templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the sets of categories, (b) a recognition component that compares an example, which is an item for which a categorization is desired, to at least one of the one or more templates to remove a category from a set of possible categories that the item is expected to belong to, and the machine outputs results of the categorization. - View Dependent Claims (19, 20, 21, 22, 23, 24, 25, 26, 27, 28)
-
-
29. A machine comprising:
-
(a) a unit that implements an evolution that builds competent elements from building blocks, wherein the competent elements are elements that are capable of distinguishing into which category of a set of categories previously uncategorized examples are to be categorized, the competent elements, being created in association with the set of categories including at least one competent element that is not itself any of the set of categories and is not a member of any of the set of categories, the set of categories being categories into which initially uncategorized examples are to be categorized, by at least computing a relationship between the competent elements and the initially uncategorized examples, wherein the initially uncategorized examples are items for which a categorization is desired, but the categorization is initially unknown, and (b) a memory that stores said competent elements, wherein the machine is configured for
receiving the initially uncategorized examples,
categorizing the initially uncategorized examples into a plurality of predetermined categories by at least computing the relationship between the competent elements and the initially uncategorized examples, and
returning an indication of the categorization.- View Dependent Claims (30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43)
-
-
44. A machine-implemented method comprising:
-
(a) storing, in a memory, a collection of templates, the collection of templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, (b) executing a learning phase which constructs said templates by at least testing a first set of templates by at least categorizing an example of an item having a known categorization into one of the set of categories based on computed relationships between each of a first set of templates and features of the item, constructing a second set of templates based on the testing and initial set of templates, and forming the collection of the templates stored in the memory based on performing the testing and the constructing one or more times. - View Dependent Claims (45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60)
-
-
61. A method comprising:
-
(a) storing in a memory a collection of templates, the collection of templates being association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, (b) executing a recognition phase that categorizes an example by at least comparing the example to at least a set of one or more templates selected from the collection of templates, wherein the example is an item for which a categorization is desired, and (c) returning the categorization. - View Dependent Claims (62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75)
-
-
76. A means for information retrieval comprising:
-
a memory which stores a collection of templates, the collection of templates being created in association with a set of categories including at least one template that is not itself any of the set of categories and is not a member of any of the set of categories, wherein the means is configured for implementing a retrieval phase which retrieves information from a set of data by at least comparing the data to fractures of said templates, and returning the information retrieved.
-
-
77. The means in 76 wherein one or more of said templates include at least a
< - DOER, ACTION>
phrase.
- DOER, ACTION>
-
78. The means in 76 wherein one or more of said templates include at least a
< - DOER, ACTION, OBJECT>
phrase.
- DOER, ACTION, OBJECT>
-
79. The means in 76 wherein one or more of said templates are concept templates.
-
80. The means in 79 wherein said concept templates are selected from the group consisting of:
-
at least one of a set of concepts occur; all of the set of concepts occur; all of the set of concepts occur in the same sentence; all of the set of concepts occur in a predetermined order; all of the set of concepts occur in the same sentence in the predetermined order; none of the set of concepts occur; at least one of the set of concepts do not occur; all of the set of concepts do not occur in the same sentence; all of the set of concepts do not occur in the predetermined order; all of the set of concepts do not occur in the same sentence in the predetermined order; all of the set of concepts occur in the same <
DOER, ACTION>
phrase;all of the set of concepts occur in the same <
DOER, ACTION, OBJECT>
phrase.
-
-
81. The means in 79 wherein the memory stores concept templates including at least
at least one of a set of concepts occur; -
all of the set of concepts occur; all of the set of concepts occur in the same sentence; all of the set of concepts occur in a predetermined order; all of the set of concepts occur in the same sentence in the predetermined order; none of the set of concepts occur; at least one of the set of concepts do not occur; all of the set of concepts do not occur in the same sentence; all of the set of concepts do not occur in the predetermined order; all of the set of concepts do not occur in the same sentence in the predetermined order; all of the set of concepts occur in the same <
DOER, ACTION>
phrase; andall of the set of concepts occur in the same <
DOER, ACTION, OBJECT>
phrase.
-
-
82. A machine comprising:
-
a storage storing a set of categories, a set of category distinguishing elements, wherein each category distinguishing element is stored in association with at least one category from the set of categories, the set of category distinguishing elements is different than the set of categories, one or more matching functions that includes at least inputs having at least a category selected from the set of categories, a category distinguishing element selected from the set of category distinguishing elements, and an example, wherein the example is an item for which a categorization is desired, an output that has at least a similarity, which is a value representing how similar the example is expected to be to the category based on the inputs; and a recognition unit that is capable of at least categorizing the example based on the set of category distinguishing elements, the matching functions, and the set of categories, the machine being configured to return results of the categorizing.
-
-
83. A system comprising a storage having stored thereon one or more category distinguishing elements, wherein the one or more category distinguishing elements are not categories, and the at least one of the one or more category distinguishing elements does not belong to any of a set of predetermined categories that the category distinguishing items are designed to distinguish and program instructions for
categorizing an uncategorized item into one of the set of predetermined categories based on the category distinguishing elements, and returning at least one of the predetermined categories based on the categorizing.
- 86. The system of 85, wherein the instructions include one or more instructions for constructing a prototype function by at least evolving an initial prototype function until the prototype function maps a set of categories to a power set of the set of values.
Specification