Large scale machine learning systems and methods
First Claim
Patent Images
1. A system, comprising:
- a repository to store a plurality of instances, each of the instances including a set of features and a label; and
at least one device to;
generate rules of a model based, at least in part, on weights and conditions formed from combinations of one or more of the features or complements of the features in the repository,identify a new instance,extract the set of features from the new instance,identify which of the rules of the model apply to the new instance based, at least in part, on the extracted features,determine a probability of the label for the new instance based, at least in part, on the weights from which the identified rules were generated, andstore information regarding the probability of the label for the new instance,where when generating the rules of the model, the at least one device is to;
select a candidate condition,determine the weight for the candidate condition,form a rule based, at least in part, on the candidate condition and the weight, andadd the rule to the model.
1 Assignment
0 Petitions
Accused Products
Abstract
A system for generating a model is provided. The system generates, or selects, candidate conditions and generates, or otherwise obtains, statistics regarding the candidate conditions. The system also forms rules based, at least in part, on the statistics and the candidate conditions and selectively adds the rules to the model.
53 Citations
24 Claims
-
1. A system, comprising:
-
a repository to store a plurality of instances, each of the instances including a set of features and a label; and at least one device to; generate rules of a model based, at least in part, on weights and conditions formed from combinations of one or more of the features or complements of the features in the repository, identify a new instance, extract the set of features from the new instance, identify which of the rules of the model apply to the new instance based, at least in part, on the extracted features, determine a probability of the label for the new instance based, at least in part, on the weights from which the identified rules were generated, and store information regarding the probability of the label for the new instance, where when generating the rules of the model, the at least one device is to; select a candidate condition, determine the weight for the candidate condition, form a rule based, at least in part, on the candidate condition and the weight, and add the rule to the model. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 14, 15, 16, 17, 18, 19, 20)
-
-
11. A method performed by one or more devices, the method comprising:
-
storing, by one or more processors associated with the one or more devices, a plurality of instances, each of the instances including a set of features and a label; generating, by one or more processors associated with the one or more devices, rules of a model based, at least in part, on conditions formed from combinations of one or more of the features associated with the stored instances or complements of the features associated with the stored instances, and on weights associated with the conditions; identifying, by one or more processors associated with the one or more devices, a new instance; identifying, by one or more processors associated with the one or more devices, the set of features from the new instance; identifying, by one or more processors associated with the one or more devices, which of the rules of the model apply to the new instance based, at least in part, on the identified features; determining, by one or more processors associated with the one or more devices, a label for the new instance based, at least in part, on the weights from which the identified rules were generated; and storing, by one or more processors associated with the one or more devices, information regarding the label for the new instance, where generating the rules of the model includes; selecting a candidate condition, determining the weight for the candidate condition, forming a rule based, at least in part, on the candidate condition and the weight, and adding the rule to the model. - View Dependent Claims (12, 13, 21)
-
-
22. One or more devices, comprising:
-
means for storing a plurality of instances, each of the instances including a set of features and a label; means for identifying a plurality of conditions as one or more of the features associated with the stored instances or a complement of one or more of the features associated with the stored instances; means for determining weights for corresponding ones of the conditions; means for forming rules for a model based, at least in part, on the conditions and the corresponding weights; means for identifying a new instance; means for identifying the set of features from the new instance; means for identifying which of the rules of the model apply based, at least in part, on the identified features; means for determining the label for the new instance based, at least in part, on the weights from which the identified rules were formed; and means for storing information regarding the label for the new instance, where the means for forming the rules for the model comprise; means for selecting a candidate condition, means for determining the weight for the candidate condition, means for forming a rule based, at least in part, on the candidate condition and the weight, and means for adding the rule to the model. - View Dependent Claims (23, 24)
-
Specification