Multistage Learner for Efficiently Boosting Large Datasets

• US 20150186795A1
• Filed: 12/30/2013
• Published: 07/02/2015
• Est. Priority Date: 12/30/2013
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Singer, Yoram, Redstone, Joshua, Chandra, Tushar Deepak, Shaked, Tal, Ie, Tze Way Eugene
• Granted Patent: US 9,418,343 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/20 Ensemble learning

Abstract

Implementations of the disclosed subject matter provide methods and systems for using a multistage learner for efficiently boosting large datasets in a machine learning system. A method may include obtaining a first plurality of examples for a machine learning system and selecting a first point in time. Next, a second point in time occurring subsequent to the first point in time may be selected. The machine learning system may be trained using m of the first plurality of examples. Each of the m examples may include a feature initially occurring after the second point in time. In addition, the machine learning system may be trained using n of the first plurality of examples, and each of the n examples may include a feature initially occurring after the first point in time.

Claims

1. A computer-implemented method comprising:

obtaining a first plurality of examples for a machine learning system;

selecting a first point in time;

selecting a second point in time occurring subsequent to the first point in time;

training the machine learning system using m of the first plurality of examples, each of the m examples including a feature initially occurring after the second point in time; and

training the machine learning system using n of the first plurality of examples, each of the n examples including a feature initially occurring after the first point in time.

2. The method of claim 1, wherein m>

n.

3. The method of claim 1, further comprising:

training the machine learning system over M iterations using only examples including a feature initially occurring after the second point in time; and

training the machine learning system over N iterations using only examples including a feature initially occurring after the first point in time.

4. The method of claim 3, wherein M>

N.

5. The method of claim 1, further comprising determining, for each of the m plurality of examples, that the example includes a feature initially occurring after the second point in time.

6. The method of claim 1, further comprising determining, for each of the n plurality of examples, that the example includes a feature initially occurring after the first point in time.

7. The method of claim 1, further comprising training the machine learning system using a second plurality of examples, wherein at least one of the second plurality includes a feature initially occurring after the first point, and at least one of the second plurality does not include any feature initially occurring after the first point.

8. The method of claim 1, further comprising selecting the first point in time based on the time at which a first feature is first present in any example obtained to train the machine learning system.

9. The method of claim 1, further comprising selecting the first point in time based on the time at which a threshold number of features are first present in any example obtained to train the machine learning system.

10. The method of claim 1, further comprising selecting the first point in time based on the time at which a threshold number of examples having a first feature are obtained to train the machine learning system.

11. The method of claim 1, further comprising selecting the first and second points in time based upon the rate at which new features are obtained to train the machine learning system.

12. A computer-implemented method comprising:

obtaining a first plurality of examples for a machine learning system;

selecting a first point in time;

selecting a second point in time occurring subsequent to the first point in time;

training the machine learning system using m of the first plurality of examples, each of the m examples including a feature initially occurring after the second point in time;

training the machine learning system using n of the first plurality of examples, each of the n examples including a feature initially occurring after the first point in time; and

training the machine learning system using a second plurality of examples, wherein at least one of the second plurality includes a feature initially occurring after the first point, and at least one of the second plurality does not include any feature initially occurring after the first point.

13. A system comprising:

a processor configured to;

obtain a first plurality of examples for a machine learning system;

select a first point in time;

select a second point in time occurring subsequent to the first point in time;

train the machine learning system using m of the first plurality of examples, each of the m examples including a feature initially occurring after the second point in time; and

train the machine learning system using n of the first plurality of examples, each of the n examples including a feature initially occurring after the first point in time.

14. The system of claim 13, wherein m>

n.

15. The system of claim 13, wherein the processor is further configured to:

train the machine learning system over M iterations using only examples including a feature initially occurring after the first point in time; and

train the machine learning system over N iterations using only examples including a feature initially occurring after the second point in time.

16. The system of claim 15, wherein M>

N.

17. The system of claim 13, wherein the processor is further configured to determine, for each of the m plurality of examples, that the example includes a feature initially occurring after the second point in time.

18. The system of claim 13, wherein the processor is further configured to determine, for each of the n plurality of examples, that the example includes a feature initially occurring after the first point in time.

19. The system of claim 13, wherein the processor is further configured to train the machine learning system using a second plurality of examples, wherein at least one of the second plurality includes a feature initially occurring after the first point, and at least one of the second plurality does not include any feature initially occurring after the first point.

20. The system of claim 13, wherein the processor is further configured to select the first point in time based on the time at which a first feature is first present in any example obtained to train the machine learning system.

21. The system of claim 13, wherein the processor is further configured to select the first point in time based on the time at which a threshold number of features are first present in any example obtained to train the machine learning system.

22. The system of claim 13, wherein the processor is further configured to select the first point in time based on the time at which a threshold number of examples having a first feature are obtained to train the machine learning system.

23. The system of claim 13, wherein the processor is further configured to select the first and second points in time based upon the rate at which new features are obtained to train the machine learning system.

24. A system comprising:

a processor configured to;

obtain a first plurality of examples for a machine learning system;

select a first point in time;

select a second point in time occurring subsequent to the first point in time;

train the machine learning system using m of the first plurality of examples, each of the m examples including a feature initially occurring after the second point in time;

train the machine learning system using n of the first plurality of examples, each of the n examples including a feature initially occurring after the first point in time; and

train the machine learning system using a second plurality of examples, wherein at least one of the second plurality includes a feature initially occurring after the first point, and at least one of the second plurality does not include any feature initially occurring after the first point.