US5155803A — Expert system for performing beta-token partition…

Claim 1

Independent Claim 576 words

A system for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality of rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions of each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. a memory configured to store the expert system; b. the memory further configured to store a set of test data elements for use in determining the satisfaction of the conditions of the rules; c. a beta-token statistics generator coupled to the memory and configured to create a testing RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once and generate processing statistics for each join node of the testing RETE network by applying the set of test data elements to the testing RETE network; d. a beta-token opinion generator coupled to the beta-token statistics generator to determine an opinion indicator based on the processing statistics for at least one of the join nodes according to the pre-selected set of heuristics; e. a beta-token decision generator coupled to the beta-token opinion generator to determine decision indicators for the at least one of the join nodes as a function of the opinion indicators for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; f. an allocator arranged to allocate the nodes of the testing RETE network into partitions as a function of the locations of the at least one of the join nodes when a positive decision indicator for the at least one of the join nodes is determined by the beta-token decision generator, each partition representing one of the multiple processors comprising the computer system; and g. a RETE network builder coupled to the allocator to convert the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 2

Independent Claim 576 words

A system for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality of rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions of each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. a memory configured to store the expert system; b. the memory further configured to store a set of test data elements for use in determining the satisfaction of the conditions of the rules; c. a beta-token statistics generator coupled to the memory and configured to create a testing RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once and generate processing statistics for each join node of the testing RETE network by applying the set of test data elements to the testing RETE network; d. a beta-token opinion generator coupled to the beta-token statistics generator to determine an opinion indicator based on the processing statistics for at least one of the join nodes according to the pre-selected set of heuristics; e. a beta-token decision generator coupled to the beta-token opinion generator to determine decision indicators for the at least one of the join nodes as a function of the opinion indicators for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; f. an allocator arranged to allocate the nodes of the testing RETE network into partitions as a function of the locations of the at least one of the join nodes when a positive decision indicator for the at least one of the join nodes is determined by the beta-token decision generator, each partition representing one of the multiple processors comprising the computer system; and g. a RETE network builder coupled to the allocator to convert the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 3

Dependent Claim 42 words

The system for building a RETE network of claim 1 wherein one of the logical connectors of the expert system represents the non-existence of a condition and that logical connector is represented by a non-existence test join node in the RETE network.

Claim 4

Dependent Claim 73 words

The system for building a RETE network of claim 2 wherein the beta-token opinion generator generates an opinion indicator for the non-existence test join node and the beta-token opinion generator is configured to modify the opinion indicator for the non-existence test join node to cause the beta-token decision generator to generate a negative decision indicator for the non-existence test join node so that beta-token partitioning never occurs at the non-existence test join node.

Claim 5

Dependent Claim 66 words

The system for building a RETE network of claim 1 wherein the RETE network is arranged to have a set of join nodes arranged as a series of linked nodes with a first join node being coupled to a subsequent join node, and a plurality of join nodes being coupled in a series to the subsequent join node, including a last join node in the series.

Claim 6

Dependent Claim 123 words

The system for building a RETE network of claim 4 wherein the join nodes of the testing RETE network have a left side input coupled to a left side memory for receiving and storing inputs and a right side input coupled to a right side memory for receiving and storing inputs, the inputs including certain ones of the test data elements, the join nodes comparing the inputs stored in the right side memory against the inputs stored in the left side memory to determine those pairs of inputs which satisfy the logical connection represented by the join node, and outputting each input pair with an indicator of the result of the comparison, as an input to other preselected nodes in the RETE network.

Claim 7

Dependent Claim 30 words

The system for building a RETE network of claim 5 wherein the beta-token statistics generator generates statistics based on each left side and right side input to each join node.

Claim 8

Dependent Claim 59 words

The system for building a RETE network of claim 6 wherein the beta-token opinion generator generates a left and a right opinion indicator corresponding to the left and right side memories of the at least one of the join nodes, respectively based on the corresponding left and right side memory inputs for the at least one of join nodes.

Claim 9

Dependent Claim 69 words

The system for building a RETE network of claim 7 wherein the beta-token opinion generator generates left and right opinion indicators for the at least one join nodes that is the first join node in the series of join nodes, and modifies the left opinion indicator of the first join node to diminish the likelihood of beta-token partitioning at that side of the first join node by the allocator.

Claim 10

Dependent Claim 88 words

The system for building a RETE network of claim 7 wherein: a. the beta-token statistic generator determines the number of comparisons that occur at the at least one of the join nodes as the set of test data elements are input to the testing RETE network; and b. the beta-token opinion generator generates opinion indicators for the at least one of the join nodes as a function of the number of comparisons at the at least one of the join nodes as determined by the beta-token statistics generator.

Claim 11

Dependent Claim 69 words

The system for building a RETE network of claim 9 wherein the beta-token opinion generator generates opinion indicators so that when the number of comparisons at the at least one of the join nodes is above a pre-determined threshold value, left and right side opinion indicators cause the beta-token decision generator to more likely generate a positive decision indicator so as to favor beta-token partitioning at the join node.

Claim 12

Dependent Claim 72 words

The system for building a RETE network of claim 9 wherein the beta-token opinion generator generates left and right side opinion indicators such that, when the number of comparisons at the at least one of the join nodes is below a pre-determined threshold value, the left and right indicators cause the beta-token decision generator to more likely generate a negative decision indicator so as to disfavor beta-token partitioning at the join node.

Claim 13

Dependent Claim 84 words

The system for building a RETE network of claim 7 wherein: a. the beta-token statistic generator determines the number of comparisons made between the left and right side memories at the first and subsequent join node as the set of test data elements are input through the testing RETE network; and b. the beta-token opinion generator generates opinion indicators for the left and right sides of the first join node as a function of the number of comparisons occurring at the subsequent join node.

Claim 14

Dependent Claim 78 words

The system for building a RETE network of claim 12 wherein the beta-token opinion generator generates opinion indicators for the left and right sides of the first join node so that when the number of comparisons occurring at the subsequent join node is above a pre-determined threshold, the indicators for the first join node causes the beta-token decision generator to more likely generate a positive decision indicator so as to favor beta-token partitioning at the first join node.

Claim 15

Dependent Claim 61 words

The system for building a RETE network of claim 7, wherein the nodes of the RETE network are arranged to receive the set of test data elements in the form of tokens, with each join node performing the comparisons corresponding to its logical connector to output data concerning the result of the comparisons to other nodes in the form of tokens.

Claim 16

Dependent Claim 72 words

The system for building a RETE network of claim 14 wherein: a. the beta-token statistic generator determines the number of tokens output at each join node of the RETE network; and b. the beta-token opinion generator generates opinion indicators for the at least one of the join nodes where the opinion indicators are generated as a function of the number of tokens output by the at least one of the join nodes.

Claim 17

Dependent Claim 66 words

The system for building an RETE network of claim 15 wherein, when the number of tokens output from the at least one of the join nodes is above a predetermined threshold, the beta-token opinion generator modifies the left and right side opinion indicators to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 18

Dependent Claim 84 words

The system to build an expert system of claim 14, wherein: a. the beta-token statistic generator determines the number of tokens received by each of the left and right side inputs of the at least one of the join nodes; b. the beta-token opinion generator generates left and right side opinion indicators for the at least one of the join nodes where the left and right side opinion indicators are determined as a function of the number of tokens received at the join node.

Claim 19

Dependent Claim 66 words

The system for building a RETE network of claim 17 wherein, when the number of tokens received at the left or right side input is above a pre-determined threshold value, the beta-token opinion generator modifies the corresponding left or right side opinion indicator to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 20

Dependent Claim 83 words

The system of building a RETE network of claim 17 wherein, when the number of tokens input to the left or right side of the subsequent join node is above a pre-determined threshold value, the corresponding left or right side opinion indicator generated for the subsequent join node by the beta-token opinion generator is modified to increase the likelihood that beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node upstream to the subsequent join node.

Claim 21

Dependent Claim 76 words

The system for building a RETE network of claim 14 wherein: a. the beta-token statistic generator determines the number of tokens output by each node of the testing RETE network; and b. the beta-token opinion generator generates left and right side opinion indicators for the at least one of the join nodes where the opinion indicators are determined as a function of the number of tokens input and tokens output to and from the join node.

Claim 22

Dependent Claim 79 words

The system for building a RETE network of claim 20 wherein, when the number of tokens received at the left or right side input of the at least one of the join nodes is greater by a pre-determined ratio than the number of tokens output from the node, the beta-token opinion generator modifies the opinion indicators to reduce the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 23

Dependent Claim 84 words

The system for building a RETE network of claim 21 wherein, when the difference between the number of tokens received by the left or right side inputs and the output of the at least one of the join nodes is greater than a predetermined threshold value, the beta-token opinion generator modifies the opinion indicator for the corresponding left or right side to reduce the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 24

Dependent Claim 84 words

The system for building a RETE network of claim 22 wherein the beta-token decision generator is arranged to generate a positive left or right side decision indicator for beta-token partitioning at the first join node when the corresponding left or right side opinion indicator for the first join node is at least as great as a pre-determined threshold value and it is at least as great as the difference between the largest opinion indicator for the subsequent join node and a pre-determined delta value.

Claim 25

Dependent Claim 70 words

The system for building a RETE network of claim 22 wherein the beta-token generator is arranged to generate a positive left or right side decision indicator for beta-token partitioning at the first join node when the corresponding left or right side opinion indicator for the first join node is equal to or greater than a pre-determined threshold value and there are no subsequent join nodes connected to that join node.

Claim 26

Dependent Claim 124 words

The system for building a RETE network of claim 22 wherein the beta-token generator is arranged to generate a positive left or right side decision indicator for beta-token partitioning at the at least one of the join nodes when the join node immediately preceding it that has negative decision indicators for beta-token partitioning, the corresponding left or right opinion indicators at the at least one of the join nodes equals or is greater than a pre-determined threshold, the at least one of the join nodes has a subsequent join node, and the largest of its opinion indicators is at least as large as the difference between the highest opinion indicator of the next subsequent join node in the series and a pre-determined delta value.

Claim 27

Dependent Claim 82 words

The system for building a RETE network of claim 7 wherein: a. the beta-token statistic generator determines processing times for each join node of the testing RETE network as each join node processes its input data; and b. the beta-token opinion generator generates opinion indicators for the left and right sides of the at least one of the join nodes where the opinion indicators are determined as a function of the processing times of the at least one of the join nodes.

Claim 28

Dependent Claim 51 words

The system for building a RETE network of claim 7 wherein the beta-token decision generator generates left and right decision indicators for the at least one of the join nodes and the allocator assigns a partition to the at least one of the nodes as a function of those decision indicators.

Claim 29

Dependent Claim 53 words

The system for building a RETE network of claim 27 wherein the beta-token decision generator determines left and right decision indicators for the at least one of the join nodes as a function of whether the corresponding left or right beta-token opinion indicators for that join node are above a pre-determined threshold value.

Claim 30

Dependent Claim 95 words

The system for building a RETE network of claim 28 wherein the beta-token decision generator modifies the left or right decision indicator for the at least one of the join nodes to be negative when the corresponding left or right opinion indicator for the at least one of the join nodes is above a pre-determined threshold value and the corresponding left or right opinion indicator of the at least one of the join nodes is also less than each opinion indicator of the next adjacent join node in the series less a pre-determined delta value.

Claim 31

Dependent Claim 57 words

The system for building a RETE network of claim 29 wherein the beta-token decision generator modifies the left or right decision indicator for the at least one of the join nodes to be negative when the corresponding left or right opinion indicator for the at least one of the join nodes is below a pre-determined threshold value.

Claim 32

Dependent Claim 72 words

The system for building a RETE network of claim 27 wherein the beta-token decision generator is arranged to generate a positive decision indicator for beta-token partitioning at the last join node when at least one of the left or right opinion indicators for the last join node is greater than or equal to a pre-determined threshold value and both the decision indicators for the adjacent join node in the series are negative.

Claim 33

Dependent Claim 56 words

The system for building a RETE network of claim 4 wherein the beta-token opinion generator generates an opinion indicator for the at least one join node that is the first in the series of join nodes and modifies the opinion indicator to diminish the likelihood of beta-token partitioning at the first join node by the allocator.

Claim 34

Dependent Claim 40 words

The system for building a RETE network of claim 1, wherein each join node is arranged to accept data inputs and the data inputs are compared by the join node in performing the test of the logical connector it represents.

Claim 35

Dependent Claim 85 words

The system for building a RETE network of claim 33 wherein: a. the beta-token statistic generator determines the number of comparisons that occur at the at least one of the join nodes as data elements are input to the testing RETE network; and b. the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes as a function of the number of comparisons at the at least one of the join nodes as determined by the beta-token statistics generator.

Claim 36

Dependent Claim 67 words

The system for building a RETE network of claim 34 wherein the beta-token opinion generator generates an opinion indicator so that when the number of comparisons at the at least one of the join nodes is above a pre-determined threshold value, the opinion indicator causes the beta-token decision generator to more likely generate a positive decision indicator so as to favor beta-token partitioning at the join node.

Claim 37

Dependent Claim 67 words

The system for building a RETE network of claim 34 wherein the beta-token opinion generator generates an opinion indicator such that, when the number of comparisons at the at least one of the join nodes is below a pre-determined threshold value, the opinion indicator causes the beta-token decision generator to more likely generate a negative decision indicator so as to disfavor beta-token partitioning at the join node.

Claim 38

Dependent Claim 131 words

The system for building a RETE network of claim 33 wherein: a. the join nodes of the testing RETE network are arranged as a series of linked nodes with the at least one of the join nodes being a first join node in the series coupled to a subsequent join node; b. the beta-token statistic generator operates to determine the number of comparisons at the first and subsequent join node as the set of test data elements are input through the testing RETE network; and c. the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes as a function of the number of comparisons occurring at the subsequent join node, wherein the at least one of the join nodes is the first join node.

Claim 39

Dependent Claim 64 words

The system for building a RETE network of claim 37 wherein the beta-token opinion generator generates an opinion indicator so that when the number of comparisons occurring at the subsequent join node is above a pre-determined threshold, the opinion indicator causes the beta-token decision generator to more likely generate a positive decision indicator so as to favor beta-token partitioning at the first join node.

Claim 40

Dependent Claim 54 words

The system for building a RETE network of claim 1, wherein the nodes of the RETE network are arranged to receive the set of test data elements in he form of tokens, with each join node performing the operation of the logical connector that it represents to output data in the form of tokens.

Claim 41

Dependent Claim 73 words

The system for building a RETE network of claim 39 wherein: a. the beta-token statistic generator determines the number of tokens output at each join node of the RETE network; and b. the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes where the opinion indicator is determined as a function of the number of tokens output by the at least one of the join nodes.

Claim 42

Dependent Claim 62 words

The system for building an RETE network of claim 40 wherein, when the number of tokens output by the at least one of the join nodes is above a predetermined threshold, the beta-token opinion generator modifies the opinion indicator to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 43

Dependent Claim 68 words

The system to build an RETE network of claim 39, wherein: a. the beta-token statistic generator determines the number of tokens received by the at least one of the join nodes; b. the beta-token opinion generator provides an opinion indicator for the at least one of the join nodes where the opinion indicator is determined as a function of the number of tokens received at the join node.

Claim 44

Dependent Claim 54 words

The system for building a RETE network of claim 42 wherein, when the number of tokens received is above a pre-determined threshold value, the beta-token opinion generator modifies the opinion indicator to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 45

Dependent Claim 41 words

The system of building a RETE network of claim 42 wherein the join nodes are arranged in a series with a first join node coupled to a subsequent join node, the first join node being upstream from the subsequent join node.

Claim 46

Dependent Claim 78 words

The system of building a RETE network of claim 44 wherein, for each join node, when the number of tokens input to the subsequent join node is above a pre-determined threshold value, the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes that increases the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node upstream to the subsequent join node.

Claim 47

Dependent Claim 69 words

The system for building a RETE network of claim 45 wherein, when the processing time for the at least one of the join nodes is above a pre-determined threshold limit, the beta-token opinion generator modifies the left and right side opinion indicators for that join node to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 48

Dependent Claim 65 words

The system for building a RETE network of claim 44 wherein, when the processing time for the at least one of the join nodes is above a pre-determined threshold limit, the beta-token opinion generator modifies the opinion indicator for that join node to increase the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 49

Dependent Claim 73 words

The system for building a RETE network of claim 39 wherein: a. the beta-token statistic generator determines the number of tokens output by each node of the testing RETE network; and b. the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes where the opinion indicator is determined as a function of the number of tokens input and tokens output to and from the join node.

Claim 50

Dependent Claim 73 words

The system for building a RETE network of claim 48 wherein, when the number of tokens received at the at least one of the join nodes is greater by a pre-determined ratio than the number of tokens output from the join node, the beta-token opinion generator modifies the opinion indicator to reduce the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 51

Dependent Claim 69 words

The system for building a RETE network of claim 49 wherein, when the difference between the number of tokens received and output by the at least one of the join nodes is greater than a pre-determined threshold value, the beta-token opinion generator modifies the opinion indicator to reduce the likelihood that the beta-token decision generator generates a positive decision indicator so that beta-token partitioning occurs at the join node.

Claim 52

Dependent Claim 69 words

The system for building a RETE network of claim 1 wherein: a. the beta-token statistic generator determines processing time for each join node of the testing RETE network as each join node processes data inputs; and b. the beta-token opinion generator generates an opinion indicator for the at least one of the join nodes where the opinion indicator is a function of the processing time for the join node.

Claim 53

Dependent Claim 21 words

The system for building a RETE network of claim 1 wherein the testing RETE network is generated on a single processor.

Claim 54

Dependent Claim 20 words

The system for building a RETE network of claim 52 wherein the beta-token statistics generator operates on the single processor.

Claim 55

Dependent Claim 20 words

The system for building a RETE network of claim 53 wherein the beta-token opinion generator operates on the single processor.

Claim 56

Dependent Claim 20 words

The system for building a RETE network of claim 54 wherein the beta-token decision generator operates on the single processor.

Claim 57

Dependent Claim 18 words

The system for building a RETE network of claim 55 wherein the allocator operates on the single processor.

Claim 58

Independent Claim 643 words

A system for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. a memory configured to store the expert system; b. the memory further configured to store a set of test data elements for use in determining the satisfaction of the conditions of the network; c. a beta-token statistics generator coupled to the memory and configured to create a test RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once and generate processing statistics for each join node of the testing RETE network by applying the set of test data elements to the RETE network; d. a beta-token opinion generator coupled to the beta-token statistics generator to determine an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; e. a beta-token decision generator coupled to the beta-token opinion generator to determine a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; f. a rule partitioner coupled to the beta-token decision generator to rule partition rules whose nodes did result in the generation of any positive beta-token decision indicators after being processed by the beta-token decision generator, the rule partitioning performed in a manner to identify conditions shared by one or more rules and partition the rules as a function of the identified shared conditions; g. an allocator arranged to allocate the nodes of the testing RETE network from the rules of the expert system into partitions based (1) on the positive decision indicators for the rules that contain nodes that have at least one positive decision indicator and (2) on the results of the rule partitioning for the rules that have no nodes that resulted in the generation of any positive decision indicators; and h. a RETE network builder coupled to the allocator to convert the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 59

Independent Claim 643 words

A system for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. a memory configured to store the expert system; b. the memory further configured to store a set of test data elements for use in determining the satisfaction of the conditions of the network; c. a beta-token statistics generator coupled to the memory and configured to create a test RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once and generate processing statistics for each join node of the testing RETE network by applying the set of test data elements to the RETE network; d. a beta-token opinion generator coupled to the beta-token statistics generator to determine an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; e. a beta-token decision generator coupled to the beta-token opinion generator to determine a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; f. a rule partitioner coupled to the beta-token decision generator to rule partition rules whose nodes did result in the generation of any positive beta-token decision indicators after being processed by the beta-token decision generator, the rule partitioning performed in a manner to identify conditions shared by one or more rules and partition the rules as a function of the identified shared conditions; g. an allocator arranged to allocate the nodes of the testing RETE network from the rules of the expert system into partitions based (1) on the positive decision indicators for the rules that contain nodes that have at least one positive decision indicator and (2) on the results of the rule partitioning for the rules that have no nodes that resulted in the generation of any positive decision indicators; and h. a RETE network builder coupled to the allocator to convert the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 60

Dependent Claim 21 words

The system for building a RETE network of claim 57 wherein the testing RETE network is generated on a single processor.

Claim 61

Dependent Claim 20 words

The system for building a RETE network of claim 58 wherein the beta-token statistics generator operates on the single processor.

Claim 62

Dependent Claim 20 words

The system for building a RETE network of claim 59 wherein the beta-token opinion generator operates on the single processor.

Claim 63

Dependent Claim 20 words

The system for building a RETE network of claim 60 wherein the beta-token decision generator operates on the single processor.

Claim 64

Dependent Claim 18 words

The system for building a RETE network of claim 61 wherein the allocator operates on the single processor.

Claim 65

Dependent Claim 19 words

The system for building a RETE network of claim 62 wherein the rule partitioner operates on the single processor.

Claim 66

Independent Claim 533 words

A method for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality of rules, each of which comprises a plurality of conditions grouped in a sequence by a set of loical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rules being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the method for building a RETE network comprising the steps of: a. storing the expert system in a memory; b. storing in the memory a set of test data elements for use in determining the satisfaction of the conditions of the rules; c. utilizing the expert system in the memory to create a testing RETE network from the stored expert system, the RETE network comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once; d. generating processing statistics for each join node of the testing RETE network by applying the set of test data elements to the RETE network; e. generating an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; f. generating a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; g. allocating the nodes of the testing RETE network into partitions based on the location of the at least one of the join nodes when a positive decision indicator for the at least one of the join nodes is generated, each partition representing one of the multiple processors comprising the computer system; and h. converting the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 67

Independent Claim 533 words

A method for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality of rules, each of which comprises a plurality of conditions grouped in a sequence by a set of loical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rules being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the conditions and operation of the logical connectors as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors of the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the method for building a RETE network comprising the steps of: a. storing the expert system in a memory; b. storing in the memory a set of test data elements for use in determining the satisfaction of the conditions of the rules; c. utilizing the expert system in the memory to create a testing RETE network from the stored expert system, the RETE network comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors are represented only once; d. generating processing statistics for each join node of the testing RETE network by applying the set of test data elements to the RETE network; e. generating an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; f. generating a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; g. allocating the nodes of the testing RETE network into partitions based on the location of the at least one of the join nodes when a positive decision indicator for the at least one of the join nodes is generated, each partition representing one of the multiple processors comprising the computer system; and h. converting the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the RETE network being arranged to execute on a corresponding processor.

Claim 68

Dependent Claim 51 words

The method for building a RETE network of claim 64 wherein one of the logical connectors of the expert system represents the non-existence of a condition and that logical connector is represented by the at least one of the join nodes as a non-existence test join node in the RETE network.

Claim 69

Dependent Claim 47 words

The method for building a RETE network of claim 65 comprising the further step of generating an opinion indicator for the non-existence test join node by modifying the opinion indicator for the non-existence test join node to prevent the beta-token partitioning of the non-existence test join node.

Claim 70

Dependent Claim 77 words

The method for building a RETE network of claim 64 wherein the RETE network is arranged to have a set of join nodes arranged as a series of linked nodes with the at least one of the join nodes being a first join node which is coupled to a subsequent join node, and a plurality of other join nodes being coupled in a series to the subsequent join node, including a last join node in the series.

Claim 71

Dependent Claim 117 words

The method for building a RETE network of claim 67 wherein the join nodes of the testing RETE network have a left side input coupled to a left side memory and a right side input coupled to a right side memory for receiving and storing inputs, the inputs including certain ones of the test data elements, the join nodes comparing the inputs stored in the right side memory against the inputs stored in the left side memory to determine those pairs of inputs which satisfy the logical connector represented by the join node, outputting each input pair with an indicator of the result of the comparison, as an input to other preselected nodes in the RETE network.

Claim 72

Dependent Claim 28 words

The method for building a RETE network of claim 68 wherein the generating step comprises the step of generating statistics based on each input to each join node.

Claim 73

Dependent Claim 43 words

The method for building a RETE network of claim 69 wherein the step of generating an opinion indicator includes the step of generating a left and a right opinion indicator for each of a left and a right side of the join node.

Claim 74

Dependent Claim 58 words

The method for building a RETE network of claim 70 comprising the further step of generating left and right opinion indicators for the first join node that diminish the likelihood of beta-token partitioning occurring at the left side of the first join node of the series when the opinion indicator is evaluated in the decision indicator generating step.

Claim 75

Dependent Claim 76 words

The method for building a RETE network of claim 70 further comprising: a. determining, in the statistics generating step, the number of comparisons that occur in each join node as the set of test data elements are input to the testing RETE network; and b. generating opinion indicators for at the least one of the join nodes as a function of the number of comparisons at the join node as determined in the statistic generating step.

Claim 76

Dependent Claim 72 words

The method for building a RETE network of claim 72 comprising the further step of modifying the left and right indicator for the at least one of the join nodes when the number of comparisons at the join node is above a predetermined threshold value so that left and right side the indicators cause generation of a decision indicator in the decision generating step to favor beta-token partitioning at the join node.

Claim 77

Dependent Claim 81 words

The method for building a RETE network of claim 72 comprising the further step of determining left and right opinion indicator corresponding to the left and right side memories in the opinion generating step such that, when the number of comparisons at the at least one of the join nodes is below a predetermined threshold value, the left and right opinion indicators cause generation of a decision indicator in the decision generating step to disfavor beta-token partitioning at the join node.

Claim 78

Dependent Claim 96 words

The method for building a RETE network of claim 70 further comprising: a. determining in the statistic generating step the number of comparisons made between the left and right side memories at the first and subsequent join nodes as data elements are input through the testing RETE network; and b. generating opinion indicators corresponding to the left and right side memories as a function of the number of comparisons occurring at the subsequent join node to cause generation of a decision indicator in the decision generating step to favor beta-token partitioning at the first join node.

Claim 79

Dependent Claim 77 words

The method for building a RETE network of claim 75 further comprising the step of generating opinion indicators corresponding to the left and right side memories of the first join node when the number of comparisons occurring at the subsequent join node is above a pre-determined threshold so that the opinion indicators for the first join node causes generation of a decision indicator in the decision generating step to favor beta-token partitioning at the first join node.

Claim 80

Dependent Claim 69 words

The method for building a RETE network of claim 70, wherein the nodes of the RETE network are arranged to receive as input certain ones of the set of test data elements in the form of tokens, performing the comparisons corresponding to the conditions and logical connectors of the expert system rule and outputting data concerning the result of the comparisons to other nodes in the form of tokens.

Claim 81

Dependent Claim 84 words

The method for building a RETE network of claim 77 comprising the further steps of: a. determining in the statistic generating step the number of tokens output at each join node of the RETE network; and b. generating in the opinion generating step left and right opinion indicators corresponding to the left and right side memories for the at least one of the join nodes where the opinion indicators are determined as a function of the number of tokens output from the join node.

Claim 82

Dependent Claim 79 words

The method for building a RETE network of claim 78 comprising the further step of modifying the left and right opinion indicators for the at least one of the join nodes in the opinion generating step such that, when the number of tokens output by the join node is above a pre-determined threshold, the left and right opinion indicators increase the likelihood that the decision indicator generated in the decision generating step favors beta-token partitioning at the join node.

Claim 83

Dependent Claim 92 words

The method to build an expert system of claim 77, comprising the further steps of: a. determining in the statistic generating step the number of tokens received by each of the left and right side inputs of each join node; b. generating modified left and right opinion indicators corresponding to the left and right side memories for the at least one of the join nodes in the opinion generating step where the left and right opinion indicators are determined as a function of the number of tokens received at the join node.

Claim 84

Dependent Claim 84 words

The method for building a RETE network of claim 80 comprising the further step of modifying the left and right opinion indicators for the at least one of the join nodes when the number of tokens received at the left or right side input is above a pre-determined threshold value, such that the corresponding left or right opinion indicator is modified to increase the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 85

Dependent Claim 98 words

The method of building a RETE network of claim 80 comprising the further step of modifying the left and right opinion indicators for the subsequent join node when the number of tokens input to the left or right side input of the subsequent join node is above a pre-determined threshold value, such that the corresponding left or right side opinion indicator generated for the subsequent join node is modified to increase the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node upstream to the subsequent join node.

Claim 86

Dependent Claim 81 words

The method for building a RETE network of claim 77 comprising the further steps of: a. determining in the statistic generating step the number of tokens output by each join node of the testing RETE network; and b. generating left and right opinion indicators for the at least one of the join nodes in the opinion generating step where the opinion indicators are determined as a function of the tokens input and of tokens output to and from the join node.

Claim 87

Dependent Claim 94 words

The method for building a RETE network of claim 83 comprising the further step of modifying the left and right opinion indicators for the at least one of the join nodes when the number of tokens received at the left or right side input of the join node is greater by a pre-determined ratio than the number of tokens output from the join node, such that the opinion indicators are modified to reduce the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 88

Dependent Claim 102 words

The method for building a RETE network of claim 84 comprising the further step of modifying the left and right opinion indicators for the at least one of the join nodes when the difference between the number of tokens received by the left or right side inputs and the output of the join node is greater than a predetermined threshold value, such that the left or right opinion indicator for the corresponding left or right side memory is modified to reduce the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 89

Dependent Claim 88 words

The method for building a RETE network of claim 70 comprising the further steps of: a. determining in the statistic generating step processing times for each join node of the test RETE network as each node processes inputs; and b. generating left and right opinion indicators corresponding to the left and right side memories of the at least one of the join nodes in the opinion generating step where the left and right opinion indicators are determined as a function of the processing times of the join node.

Claim 90

Dependent Claim 52 words

The method for building a RETE network of claim 70 wherein the decision generating step determines left and right decision indicators for the at least one of the join nodes as a function of whether the corresponding left or right opinion indicator for the join node is above a pre-determined threshold value.

Claim 91

Dependent Claim 90 words

The method for building a RETE network of claim 87 comprising the further step of modifying in the decision generating step the left or right decision indicator for the at least one of the join nodes to be negative when the corresponding left or right opinion indicator for the join node is above a predetermined threshold value and the corresponding left or right opinion indicator of the join node is also less than each opinion indicator of the next subsequent join node in the series less a pre-determined delta value.

Claim 92

Dependent Claim 57 words

The method for building a RETE network of claim 87 comprising the further step of modifying in the decision generating step the left or right decision indicator for the at least one of the join nodes to be negative when the corresponding left or right opinion indicator for the join node is below a predetermined threshold value.

Claim 93

Dependent Claim 74 words

The method for building a RETE network of claim 70 comprising the further step of generating in the decision generating step a positive decision indicator for beta-token partitioning at the last join node when at least one of the left or right opinion indicators for the last join node is greater than or equal to a predetermined threshold value and both the decision indicators for the adjacent join node in the series are negative.

Claim 94

Dependent Claim 85 words

The method for building a RETE network of claim 70 comprising the further step of generating in the decision generating step a positive left or right side decision indicator for beta-token partitioning at the first join node when the corresponding left or right side opinion indicator for the first join node is at least as great as a pre-determined threshold value and is at least as great as the difference between the largest opinion indicator for the subsequent join node and a predetermined delta value.

Claim 95

Dependent Claim 75 words

The method for building a RETE network of claim 70 comprising the further step of generating in the decision generating step a positive left or right side decision indicator for beta-token partitioning the first join node when the corresponding left or right side opinion indicator for the first join node is equal to or greater than a pre-determined threshold value and there are no subsequent join nodes connected to the join node in the series.

Claim 96

Dependent Claim 127 words

The method for building a RETE network of claim 70 comprising the further step of generating in the decision generating step a positive left or right side decision indicator for beta-token partitioning at the at least one of the join nodes when the join node immediately preceding it that has negative decision indicators for beta-token partitioning, the corresponding left or right opinion indicators at the at least one of the join nodes equals or exceeds a predetermined threshold, the at least one of the join nodes has a subsequent join node in the series, and the largest of its opinion indicators is at least as large as the difference between the highest opinion indicator of the subsequent join node in the series and a pre-determined delta value.

Claim 97

Dependent Claim 67 words

The method for building a RETE network of claim 67 comprising the further step of generating an opinion indicator for the first join node by modifying the opinion indicator for the join node so that the opinion indicator diminishes the likelihood of beta-token partitioning at the first join node in the series for each rule, when the opinion indicator is evaluated in the decision indicator generating step.

Claim 98

Dependent Claim 42 words

The method for building a RETE network of claim 64, wherein each join node is arranged to accept data elements as inputs and the data elements are compared to determine which data elements satisfy the logical connector represented by the join node.

Claim 99

Dependent Claim 77 words

The method for building a RETE network of claim 95 further comprising: a. determining, in the statistics generating step, the number of comparisons that occur at each join node as the set of test data elements are input to the test RETE network; and b. generating an opinion indicator for the at least one of the join nodes as a function of the number of comparisons at the join node as determined in the statistic generating step.

Claim 100

Dependent Claim 63 words

The method for building a RETE network of claim 96 comprising the further step of modifying the opinion indicator when the number of comparisons at the at least one of the join nodes is above a pre-determined threshold value so that the opinion indicator causes generation of a decision indicator in the decision generating step to favor beta-token partitioning at the join node.

Claim 101

Dependent Claim 68 words

The method for building a RETE network of claim 96 comprising the further step of determining an opinion indicator in the opinion generating step such that, when the number of comparisons at the at least one of the join nodes is below a pre-determined threshold value, the opinion indicator causes generation of a decision indicator in the decision generating step to disfavor beta-token partitioning at the join node.

Claim 102

Dependent Claim 156 words

The method for building a RETE network of claim 95 comprising the further step of arranging the testing RETE network as a series of linked nodes with a first join node in the series being coupled to a subsequent join node, the method comprising the further steps of: a. determining in the statistic generating step the number of comparisons occurring at the first and subsequent join nodes as data elements are input through the testing RETE network; and b. generating an opinion indicator for the at least one of the join nodes as a function of the number of comparisons occurring at the subsequent join node to cause generation of a decision indicator for the at least one of the join nodes in the decision generating step to favor beta-token partitioning at the at least one of the join nodes, wherein the at the at least one of the join nodes is the first join node.

Claim 103

Dependent Claim 75 words

The method for building a RETE network of claim 99 further comprising the step of generating an opinion indicator for the first join node, in the opinion generating step, when the number of comparisons occurring at the subsequent join node is above a pre-determined threshold so that the opinion indicator causes generation of a decision indicator for the first join node in the decision generating step to favor beta-token partitioning at the first join node.

Claim 104

Dependent Claim 69 words

The method for building a RETE network of claim 64, wherein the nodes of the RETE network are arranged to receive as input certain ones of the set of test data elements in the form of tokens, performing the comparisons corresponding to the conditions and logical connectors of the expert system rule and outputting data concerning the result of the comparisons to other nodes in the form of tokens.

Claim 105

Dependent Claim 74 words

The method for building a RETE network of claim 101 comprising the further steps of: a. determining in the statistics generating step the number of tokens output at each join node of the RETE network; and b. generating in the opinion generating step an opinion indicator for the at least one of the join nodes where the opinion indicator is determined as a function of the number of tokens output from the join node.

Claim 106

Dependent Claim 74 words

The method for building an RETE network of claim 102 comprising the further step of modifying the opinion indicator for the at least one of the join nodes in the opinion generating step such that, when the number of tokens output by the join node is above a pre-determined threshold, the opinion indicator increases the likelihood that the decision indicator generated in the decision generating step favors beta-token partitioning occurs at the join node.

Claim 107

Dependent Claim 65 words

The method to build a RETE network of claim 101, wherein: a. determining in the statistic generating step the number of tokens received by each join node; b. generating in the opinion generating step an opinion indicator for the at least one of the join nodes where the opinion indicator is determined as a function of the number of tokens received at the join node.

Claim 108

Dependent Claim 61 words

The method for building a RETE network of claim 104 comprising the further step of modifying the opinion indicator when the number of tokens received is above a pre-determined threshold value, such that the opinion indicator is modified to increase the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 109

Dependent Claim 41 words

The method of building a RETE network of claim 104 wherein the join nodes are arranged in a series with a first join node coupled to a subsequent join node, the first join node being upstream from the subsequent join node.

Claim 110

Dependent Claim 84 words

The method of building a RETE network of claim 106 wherein the at least one of the join nodes is the first join node and wherein, when the number of tokens input to the subsequent join node is above a pre-determined threshold value, the opinion indicator generated in the opinion generating step for first join node increases the likelihood that the decision indicator for the first join node generated in the decision generation step determines that beta-token partitioning occurs at the first join node.

Claim 111

Dependent Claim 83 words

The method for building a RETE network of claim 107 comprising the further step of modifying the left and right opinion indicators for the at least one of the join nodes when the processing time for the join node is above a pre-determined threshold limit, such that the left and right side opinion indicators for the join node are modified to increase the likelihood that the decision indicator generated in the decision generation step determines that beta-token partitioning occurs at the join node.

Claim 112

Dependent Claim 74 words

The method for building a RETE network of claim 106 comprising the further step of modifying the opinion indicator for the at least one of the join nodes when the processing time for the join node is above a pre-determined threshold limit, such that the opinion indicator for the join node is modified to increase the likelihood that the decision in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 113

Dependent Claim 81 words

The method for building a RETE network of claim 101 comprising the further steps of: a. determining in the statistic generating step the number of tokens output by each join node of the testing RETE network; and b. generating an opinion indicator for the at least one of the join nodes in the opinion generating step where the opinion indicator is determined as a function of the number of tokens input and of tokens output to and from the join node.

Claim 114

Dependent Claim 84 words

The method for building a RETE network of claim 110 comprising the further step of modifying the opinion indicator for the at least one of the join nodes when the number of tokens received at the join node is greater by a predetermined ratio than the number of tokens output from the join node, such that the opinion indicator is modified to reduce the likelihood that a decision indicator generator in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 115

Dependent Claim 80 words

The method for building a RETE network of claim 111 comprising the further step of modifying the opinion indicator for the at least one of the join nodes when the difference between the number of tokens received and output by the join node is greater than a pre-determined threshold value, such that the opinion indicator is modified to reduce the likelihood that a decision indicator generated in the decision generating step determines that beta-token partitioning occurs at the join node.

Claim 116

Dependent Claim 73 words

The method for building a RETE network of claim 64 comprising the further steps of: a. determining in the statistic generating step processing times for each join node of the testing RETE network as each node processes inputs; and b. generating an opinion indicator in the opinion generating step for the at least one of the join nodes where the opinion indicator is a function of the processing time for the join node.

Claim 117

Dependent Claim 25 words

The method of building a RETE network of claim 64 wherein the step of generating the testing RETE network is performed on a single processor.

Claim 118

Dependent Claim 21 words

The method for building a RETE network of claim 114 wherein the statistic generating step is performed on the single processor.

Claim 119

Dependent Claim 21 words

The method for building a RETE network of claim 115 wherein the opinion generating step is performed on the single processor.

Claim 120

Dependent Claim 21 words

The method for building a RETE network of claim 116 wherein the decision generating step is performed on the single processor.

Claim 121

Dependent Claim 20 words

The method for building a RETE network of claim 117 wherein the allocating step is performed on the single processor.

Claim 122

Independent Claim 609 words

A method for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors that comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the condition and operation of the logical operation as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors found in the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. storing the expert system in a memory; b. storing a set of test data elements for use in determining the satisfaction of the conditions of the network in the memory; c. accessing the expert system and data elements in the memory to create a testing RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors represented only once; d. generating processing statistics for each join node of the testing RETE network by applying the set of test data elements to the network; e. generating an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; f. generating a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; g. partitioning the rules of the expert system according to a rule partition scheme to rule partition rules whose nodes did not result in the generation of any positive beta-token decision indicators after being processed in the decision generating step, the rule partitioning performed in a manner to identify conditions shared by one or more rules and partition the rules as a function of the identified shared conditions; h. allocating the nodes of the testing RETE network from the rules of the expert system into partitions based (1) on the positive decision indicators for the rules that contain nodes that have at least one positive decision indicator and (2) on the results of the rule partitioning step for the rules that did not result in the generation of positive decision indicators; and i. converting the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the testing RETE network being arranged to execute on a corresponding processor.

Claim 123

Independent Claim 609 words

A method for building a reticular discrimination network (RETE network) capable of being partitioned across a set of multiple, interconnected computer processors that comprising a computer system, from at least a portion of a target expert system, the target expert system comprising a plurality rules each of which comprises a plurality of conditions grouped in a sequence by a set of logical connectors and a result to occur when the conditions and the logical connectors of the rule are satisfied, the conditions and logical connectors of the rule being satisfied in comparison against a set of data elements, certain ones of the plurality of rules having conditions that are equivalent to conditions in other ones of the plurality of rules, certain ones of the plurality of rules having sequences of conditions and logical connectors that are equivalent to sequences of conditions and logical connectors in other ones of the plurality of the rules, each of the conditions in the a rule being represented by a test node and each of the logical connectors being represented by a join node in the RETE network, the test and join nodes being arranged to determine the satisfaction of the condition and operation of the logical operation as represented by the nodes, the nodes also being coupled together to represent the sequence of conditions and logical connectors found in the rules of the expert system, the functions each of the nodes capable of being performed on one of the multiple processors of the computer system, the system for building a RETE network comprising: a. storing the expert system in a memory; b. storing a set of test data elements for use in determining the satisfaction of the conditions of the network in the memory; c. accessing the expert system and data elements in the memory to create a testing RETE-net from the stored expert system comprising a set of test nodes and join nodes representing the conditions and logical connectors in the rules of the expert system arranged such that shared sequences of conditions and logical connectors represented only once; d. generating processing statistics for each join node of the testing RETE network by applying the set of test data elements to the network; e. generating an opinion indicator based on the processing statistics for at least one of the join nodes according to a pre-selected set of heuristics; f. generating a decision indicator for the at least one of the join nodes as a function of the opinion indicator for the at least one of the join nodes, the decision indicators being positive or negative indicators that are based on a pre-determined set of decision making heuristics; g. partitioning the rules of the expert system according to a rule partition scheme to rule partition rules whose nodes did not result in the generation of any positive beta-token decision indicators after being processed in the decision generating step, the rule partitioning performed in a manner to identify conditions shared by one or more rules and partition the rules as a function of the identified shared conditions; h. allocating the nodes of the testing RETE network from the rules of the expert system into partitions based (1) on the positive decision indicators for the rules that contain nodes that have at least one positive decision indicator and (2) on the results of the rule partitioning step for the rules that did not result in the generation of positive decision indicators; and i. converting the rules of the target expert system into a computer-executable RETE network, according to the partitions, each partition of the testing RETE network being arranged to execute on a corresponding processor.

Claim 124

Dependent Claim 25 words

The method for building a RETE network of claim 117 wherein the step of generating the test RETE network is performed on a single processor.

Claim 125

Dependent Claim 21 words

The method for building a RETE network of claim 120 wherein the statistic generating step is performed on the single processor.

Claim 126

Dependent Claim 21 words

The method for building a RETE network of claim 121 wherein the opinion generating step is performed on the single processor.

Claim 127

Dependent Claim 21 words

The method for building a RETE network of claim 122 wherein the decision generating step is performed on the single processor.

Claim 128

Dependent Claim 20 words

The method for building a RETE network of claim 123 wherein the allocating step is performed on the single processor.

Claim 129

Dependent Claim 21 words

The method for building a RETE network of claim 124 wherein the rule partitioning step is performed on the single processor.

Annuity Stage	Window Opens	Payment Deadline	Surcharge Period Until	Estimated Fee	Status
3.5 Years (Year 4)	1995-10-13	1996-04-13	1996-10-13	$2000.00	Window Closed / Lapsed
7.5 Years (Year 8)	1999-10-13	2000-04-13	2000-10-13	$3760.00	Window Closed / Lapsed
11.5 Years (Year 12)	2003-10-13	2004-04-13	2004-10-13	$7700.00	Window Closed / Lapsed

Expert system for performing beta-token partitioning in a rete network

Abstract

Inventors

Bibliographic Details

Claims Analysis

References Cited (Backward Citations)

USPTO Annuity & Maintenance Windows

Assigned Categories & Tags

Assign Category Tag