Processing dynamic data within an adaptive oracle-trained learning system using dynamic data set distribution optimization

Patent metadata

Field	Value
Publication number	US-2022180250-A1
Application number	US-202117528514-A
Country	US
Kind code	A1
Filing date	Nov 17, 2021
Priority date	Dec 23, 2013
Publication date	Jun 9, 2022
Grant date	—

Abstract

In general, embodiments of the present invention provide systems, methods and computer readable media for an adaptive oracle-trained learning framework for automatically building and maintaining models that are developed using machine learning algorithms. In embodiments, the framework leverages at least one oracle (e.g., a crowd) for automatic generation of high-quality training data to use in deriving a model. Once a model is trained, the framework monitors the performance of the model and, in embodiments, leverages active learning and the oracle to generate feedback about the changing data for modifying training data sets while maintaining data quality to enable incremental adaptation of the model.

First claim

1 - 20 . (canceled)

21 . A system, comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to:
determine feature data indicative of a set of features associated with a training data set for a predictive model; determine, based on the feature data associated with the training data set, a training distribution representative of a goal for the predictive model; apply an input data instance to the predictive model to determine a label for the input data instance; add the input data instance to a data bin of a data reservoir based on the label, wherein the data reservoir comprises candidate training data for the training data set; in response to a determination that the training distribution representative of the goal for the predictive model comprises goal criteria associated with at least one label that corresponds to at least the label, select at least a portion of the candidate training data from at least the data bin of the data reservoir that comprises the input data instance; train the predictive model based on at least the portion of the candidate training data to generate an updated predictive model; and classify data collected from one or more network sources based on the updated predictive model.

22 . The system of claim 21 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
determine a confidence value associated with the label for the input data instance; and add the input data instance to the data bin of the data reservoir in response to the confidence value satisfying a defined confidence value threshold.

23 . The system of claim 21 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
configure a set of data bins of the data reservoir based on the training distribution representative of the goal for the predictive model.

24 . The system of claim 21 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
configure a set of data bins of the data reservoir based on configuration data related to size capacity for respective data bins.

25 . The system of claim 24 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
add the input data instance to a data bin of a data reservoir in response to the data bin satisfying size capacity criterion.

26 . The system of claim 24 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
replace a particular input data instance stored in the data bin of the data reservoir with the input data instance in response to the data bin not satisfying size capacity criterion.

27 . The system of claim 21 , wherein the one or more storage devices store instructions that are operable, when executed by the one or more computers, to further cause the one or more computers to:
allocate respective labels for respective data bins of the data reservoir; and compare the label for the input data instance to the respective labels for the respective data bins of the data reservoir.

28 . A computer-implemented method, comprising:
determining, by a computing device comprising a processor, feature data indicative of a set of features associated with a training data set for a predictive model; determining, by the computing device and based on the feature data associated with the training data set, a training distribution representative of a goal for the predictive model; applying, by the computing device, an input data instance to the predictive model to determine a label for the input data instance; adding, by the computing device, the input data instance to a data bin of a data reservoir based on the label, wherein the data reservoir comprises candidate training data for the training data set; in response to a determination that the training distribution representative of the goal for the predictive model comprises goal criteria associated with at least one label that corresponds to at least the label, selecting, by the computing device, at least a portion of the candidate training data from at least the data bin of the data reservoir that comprises the input data instance; training, by the computing device, the predictive model based on at least the portion of the candidate training data to generate an updated predictive model; and classifying, by the computing device, data collected from one or more network sources based on the updated predictive model.

29 . The computer-implemented method of claim 28 , further comprising:
determining, by the computing device, a confidence value associated with the label for the input data instance; and adding, by the computing device, the input data instance to the data bin of the data reservoir in response to the confidence value satisfying a defined confidence value threshold.

30 . The computer-implemented method of claim 28 , further comprising:
configuring, by the computing device, a set of data bins of the data reservoir based on the training distribution representative of the goal for the predictive model.

31 . The computer-implemented method of claim 28 , further comprising:
configuring, by the computing device, a set of data bins of the data reservoir based on configuration data related to size capacity for respective data bins.

32 . The computer-implemented method of claim 31 , further comprising:
adding, by the computing device, the input data instance to a data bin of a data reservoir in response to the data bin satisfying size capacity criterion.

33 . The computer-implemented method of claim 31 , further comprising:
replacing, by the computing device, a particular input data instance stored in the data bin of the data reservoir with the input data instance in response to the data bin not satisfying size capacity criterion.

34 . The computer-implemented method of claim 28 , further comprising:
allocating, by the computing device, respective labels for respective data bins of the data reservoir; and comparing, by the computing device, the label for the input data instance to the respective labels for the respective data bins of the data reservoir.

35 . A computer program product, stored on a computer readable medium, comprising instructions that when executed by one or more computers cause the one or more computers to:
determine feature data indicative of a set of features associated with a training data set for a predictive model; determine, based on the feature data associated with the training data set, a training distribution representative of a goal for the predictive model; apply an input data instance to the predictive model to determine a label for the input data instance; add the input data instance to a data bin of a data reservoir based on the label, wherein the data reservoir comprises candidate training data for the training data set; in response to a determination that the training distribution representative of the goal for the predictive model comprises goal criteria associated with at least one label that corresponds to at least the label, select at least

Assignees

• Groupon Inc

Inventors

• Jeffery Shawn Ryan
• Johnston David Alan

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/04

  Inference or reasoning models · CPC title