Back to results
US 11,455,551 B2Granted
Data blaming
AssigneePalo Alto Networks, Inc.
InventorsWilliam Redington Hewlett, II, Seokkyung Chung, Lin Xu
Life of this patent
Estimating the term…
Abstract
An identification of an item that was misclassified by a classification model constructed in accordance with a machine learning technique is received. One example of such a machine learning technique is a random forest. A subset of training data, previously used to construct the model, and that is associated with the misclassified item is identified. At least a portion of the identified subset is provided as output.
Claims21
A system, comprising:
- a processor configured to:
- receive an identification of an item that was misclassified by a classification model constructed in accordance with a machine learning technique, and in response:
- identify a subset of a training data set used to construct the classification model, wherein the subset comprises at least one training sample, for which inclusion of the at least one training sample in the training data set during the construction of the classification model is determined to have likely contributed to a misclassification of the misclassified item, and wherein identifying the subset of the training data set includes creating a prioritized list of training samples sorted by a count of how often a given training sample shares a leaf with a misclassified item;
- update the training data set in response to the identification in the subset of the at least one training sample to form an updated training set;
- and
- generate an updated classification model using the updated training set;
- a memory coupled to the processor and configured to provide the processor with instructions.
4 more elements
A method, comprising:
- receiving an identification of an item that was misclassified by a classification model constructed in accordance with a machine learning technique, and in response:
- identifying a subset of a training data set used to construct the classification model, wherein the subset comprises at least one training sample, for which inclusion of the at least one training sample in the training data set during the construction of the classification model is determined to have likely contributed to a misclassification of the misclassified item, and wherein identifying the subset of the training data set includes creating a prioritized list of training samples sorted by a count of how often a given training sample shares a leaf with a misclassified item;
- updating the training data set in response to the identification in the subset of the at least one training sample to form an updated training set;
- and
- generating an updated classification model using the updated training set.
2 more elements
A computer program product embodied in a non-transitory computer readable storage medium and comprising computer instructions for:
- receiving an identification of an item that was misclassified by a classification model constructed in accordance with a machine learning technique, and in response:
- identifying a subset of a training data set used to construct the classification model, wherein the subset comprises at least one training sample, for which inclusion of the at least one training sample in the training data set during the construction of the classification model is determined to have likely contributed to a misclassification of the misclassified item, and wherein identifying the subset of the training data set includes creating a prioritized list of training samples sorted by a count of how often a given training sample shares a leaf with a misclassified item;
- updating the training data set in response to the identification in the subset of the at least one training sample to form an updated training set;
- and
- generating an updated classification model using the updated training set.
2 more elements
Dependencies are inferred from the claim text.