Updating logistic regression models using coherent gradient
AssigneeProofpoint, Inc.
InventorGordon P. Rios
Life of this patent
Estimating the term…
Abstract
A document classification system updates the logistic regression parameters of the logistic regression algorithm used to classify the documents, using a coherent gradient method based upon the coherency between the update data regarding the classifications of the documents and the combined data including both the update data and the original data used to generate the original logistic regression parameters. The classification system can be applied to an email filtering system for filtering spam email.
Claims30
A computer system for classifying documents using a logistic regression algorithm, the computer system comprising:
- a computer readable storage medium storing computer instructions including:
- a logistic regression engine for applying the logistic regression algorithm to the documents to classify the documents;
- and
- a logistic regression model updating module for generating second logistic regression parameters updating first logistic regression parameters of the logistic regression algorithm, the first logistic regression parameters being generated based upon a first set of data, and the second logistic regression parameters being generated based upon a second set of data if the second set of data is coherent with a third set of data and the second logistic regression parameters not being generated based upon the second set of data if the second set of data is not coherent with the third set of data, the third set of data including external classifications of the documents and the second set of data including a combination of the first and the third sets of data, and wherein the second set of data is coherent with the third set of data if first and second changes to the first logistic regression parameters to generate the second logistic regression parameters, as indicated by the second and third sets of data, respectively, are consistent with each other, the logistic regression engine applying the logistic regression algorithm to the documents to classify the documents using the second logistic regression parameters if the second logistic regression parameters are generated based upon the second set of data and using the first logistic regression parameters if the second logistic regression parameters are not generated based upon the second set of data;
- and
- a processor for executing the computer instructions.
A computer-implemented method for generating second logistic regression parameters updating first logistic regression parameters of a logistic regression algorithm used in classifying documents, the first logistic regression parameters being generated based upon a first set of data and the second logistic regression parameters being generated based upon a second set of data, the computer-implemented method configured as computer instructions stored on a computer readable storage medium to cause a processor of a computer to execute the computer instructions, the computer-implemented method comprising:
- receiving a third set of data, the third set of data including external classifications of the documents;
- and
- generating the second logistic regression parameters based upon the second set of data if the second set of data is coherent with the third set of data and not generating the second logistic regression parameters based upon the second set of data if the second set of data is not coherent with the third set of data, the second set of data including a combination of the first and the third sets of data, the first logistic regression parameters being generated based upon the first set of data, and wherein the second set of data is coherent with the third set of data if first and second changes to the first logistic regression parameters for generating the second logistic regression parameters, as indicated by the second and third sets of data, respectively, are consistent with each other, and the logistic regression algorithm being applied to the documents to classify the documents using the second logistic regression parameters if the second logistic regression parameters are generated based upon the second set of data and using the first logistic regression parameters if the second logistic regression parameters are not generated based upon the second set of data.
A computer readable storage medium storing a computer program product including computer instructions configured to cause a processor of a computer to perform a method for generating second logistic regression parameters updating first logistic regression parameters of a logistic regression algorithm used in classifying documents, the first logistic regression parameters being generated based upon a first set of data and the second logistic regression parameters being generated based upon a second set of data, the method comprising:
- receiving a third set of data, the third set of data including external classifications of the documents;
- and
- generating the second logistic regression parameters based upon the second set of data if the second set of data is coherent with the third set of data and not generating the second logistic regression parameters based upon the second set of data if the second set of data is not coherent with the third set of data, the second set of data including a combination of the first and the third sets of data, the first logistic regression parameters being generated based upon the first set of data, and wherein the second set of data is coherent with the third set of data if first and second changes to the first logistic regression parameters for generating the second logistic regression parameters, as indicated by the second and third sets of data, respectively, are consistent with each other, and the logistic regression algorithm being applied to the documents to classify the documents using the second logistic regression parameters if the second logistic regression parameters are generated based upon the second set of data and using the first logistic regression parameters if the second logistic regression parameters are not generated based upon the second set of data.
Dependencies are inferred from the claim text.