Systems and methods for unsupervised feature selection for online machine learning
AssigneeActimize Ltd.
InventorDanny BUTVINIK
Life of this patent
Estimating the term…
Abstract
Systems and methods for unsupervised feature selection for online machine learning are provided. Features can be selected from a plurality of online data sources having a plurality of respective online data streams, and an aggregated feature set and aggregated data can be formed therefrom. The aggregated feature set and the aggregated data can be used by machine learning models in real time to provide real time online machine learning.
Claims19
A method for unsupervised feature selection for online machine learning, the method comprising:
- receiving, via a computer, a first online data stream from a first online data source and a second online data stream from a second online data source;
- at a periodic frequency, determining, via the computer, a first set of features based on a portion of the first online data stream, where the portion of the first online data stream changes each period;
- at the periodic frequency, determining, via the computer, a second set of features based on a portion of the second data stream, where the portion of the second data stream changes each period;
- and
- at the periodic frequency, determining, via the computer, an aggregate feature set for updating one or more online machine learning models in real time, such that the one or more one line machine learning models are usable in real time.
A system for unsupervised feature selection for online machine learning, the systems comprising:
- a plurality of online data sources that each respectively transmit an online data stream;
- a plurality of processors, each in communication with one of the plurality of online data sources, each configured to:
- receive an online data stream from the particular online data source the particular processor is in communication with,
- determine a set of features based on a portion of the received online data stream;
- and
- determine an aggregate feature set for updating one or more online machine learning models in real time, such that the one or more one line machine learning models are usable in real time.
The system of claim 9 wherein the portion of the online data stream is based on a size of a buffer in the computer.
A non-transitory computer readable storage medium storing a set of instructions for causing a computer to unsupervised feature selection for online machine learning, by performing the operations of:
- receiving a first online data stream from a first online data source and a second online data stream from a second online data source;
- at a periodic frequency, determining a first set of features based on a portion of the first online data stream, where the portion of the first online data stream changes each period;
- at the periodic frequency, determining a second set of features based on a portion of the second data stream, where the portion of the second data stream changes each period;
- and
- at the periodic frequency, determining an aggregate feature set for updating one or more online machine learning models in real time, such that the one or more one line machine learning models are usable in real time.
Dependencies are inferred from the claim text.