Resilient classification of data
AssigneeMicrosoft Corporation
InventorsSrivatsan Laxman, Ramarathnam Venkatesan
Life of this patent
Estimating the term…
Abstract
A classification system is described for resilient classification of data. In various embodiments, the classification system divides a full set of the training data into a positive subset and a negative subset. The positive subset contains all training data with a positive classification value and the negative subset contains all training data with a negative classification value. The classification system constructs several subsets of the training data wherein each subset contains data randomly selected from both the positive subset and the negative subset. The classification system then creates at least two classifiers based on each of the randomly created subsets of the training data.
Claims20
A method performed by a computer system for classifying data with a classification system when training data that is used to train the classification system potentially contains errors, comprising:
- dividing a full set of the training data into a positive subset and a negative subset, the positive subset containing all training data with a positive classification value and the negative subset containing all training data with a negative classification value;
- creating at least two randomly constructed subsets of the full set of training data wherein each randomly created subset contains data selected with equal probability from both the positive subset and the negative subset further wherein a count of data in each randomly created subset is less than a count of the full set of training data;
- and
- creating at least two classifiers based on each of the randomly created subsets of the training data.
A system for resiliently classifying data, comprising:
- a dividing component that divides a full set of training data into a positive subset and a negative subset, the positive subset containing all training data with a positive classification value and the negative subset containing all training data with a negative classification value, and constructs at least two subsets of data randomly selected from the full set of training data wherein each subset contains data selected with a specified probability from the positive and the negative subsets, further wherein a count of data in each constructed subset is less than a count of the full set of training data;
- and
- at least two support vector machines, each support vector machine based on one of the constructed subsets of the training data.
A computer-readable medium storing computer-executable instructions that, when executed, cause a computer system to perform a method for resiliently classifying data, the method comprising:
- dividing a portion of training data into a positive subset and a negative subset, the positive subset containing all training data with a positive classification value and the negative subset containing all training data with a negative classification value;
- constructing at least two subsets of the portion of the training data wherein each constructed subset contains data that is randomly selected from both the positive subset and the negative subset;
- and
- creating at least two support vector machines based on each of the randomly created subsets of the training data.
Dependencies are inferred from the claim text.