Almost online large scale collaborative filtering based recommendation system

• US 9,348,924 B2
• Filed: 03/15/2013
• Issued: 05/24/2016
• Est. Priority Date: 03/15/2013
• Status: Active Grant

• Current Assignee: R2 Solutions LLC (Acacia Research Corporation)
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Somekh, Oren Shlomo, Golbandi, Nadav, Rokhlenko, Oleg, Lempel, Ronny
• Primary Examiner(s): Richardson, James E
• Application Number:
  US14/123,321
• Publication Number:
  US 20140280251A1
• Time in Patent Office: 1,166 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/0269 based on user profile or attribute

Abstract

A method for adjusting one or more parameters associated with a model. The method comprises obtaining, from a first source, first information related to activity of a user. The method further comprises adjusting one or more parameters associated with a model based on the first information collected within a first length of time, and obtaining, from a second source, second information related to activity of the user. The method further comprises adjusting the one or more parameters associated with the model based on the second information collected within a second length of time and a measure indicative of performance of the model, wherein the second length of time is larger than the first length of time.

Claims

1. A method implemented on a machine having at least one processor, a storage, and a communication platform for adjusting one or more parameters associated with a model, comprising:

obtaining, from a first source, first information related to activity of a user;

adjusting one or more parameters associated with a model based on the first information obtained within a first time period having a first length of time;

obtaining, from a second source, second information related to activity of the user;

adjusting at least the one or more parameters associated with the model based on the second information obtained within a second time period having a second length of time and a measure indicative of performance of the model;

changing the first length of time when the adjustment of the one or more parameters based on the first information exceeds a first threshold; and

changing the second length of time when the adjustment of the at least one or more parameters based on the second information exceeds a second threshold, wherein

the model is used to determine an affiliation between the user and content, and

the second length of time is larger than the first length of time.

2. The method of claim 1, wherein the second time period overlaps with the first time period.

3. The method of claim 1, wherein the step of adjusting one or more parameters associated with the model based on the first information includes:

performing an incremental update of values of the one or more parameters based on the first information.

4. The method of claim 1, wherein the step of adjusting at least the one or more parameters associated with the model based on the second information includes:

training the model using a collaborative filtering approach based on the second information.

5. The method of claim 1, wherein the affiliation is based on a score computed based on the model and bias with respect to the user and the content.

6. The method of claim 1, wherein the affiliation is based on a score computed based on the model and latent factor vectors with respect to the user and the content.

7. A system having at least one processor for adjusting one or more parameters associated with a model, the system comprising:

a modeling enhancer implemented on the at least one processor and configured to obtain, from a first source, first information related to activity of a user, and obtain, from a second source, second information related to activity of the user;

a first adjuster implemented on the at least one processor and configured to adjust one or more parameters associated with a model based on the first information obtained within a first time period having a first length of time;

a second adjuster implemented on the at least one processor and configured to adjust at least the one or more parameters associated with the model based on the second information obtained within a second time period having a second length of time and a measure indicative of performance of the model;

a short term length adjuster configured to change the first length of time when the adjustment of the one or more parameters based on the first information exceeds a first threshold; and

a long term length adjuster configured to change the second length of time when the adjustment of the at least one or more parameters based on the second information exceeds a second threshold, wherein

the model is used to determine an affiliation between the user and content, and

the second length of time is larger than the first length of time.

8. The system of claim 7, wherein the second time period overlaps with the first time period.

9. The system of claim 7, wherein the first adjuster is further configured to perform an incremental update of values of the one or more parameters based on the first information.

10. The system of claim 7, wherein the second adjuster is further configured to train the model using a collaborative filtering approach based on the second information.

11. The system of claim 7, wherein the affiliation is based on a score computed based on the model and bias with respect to the user and the content.

12. The system of claim 7, wherein the affiliation is based on a score computed based on the model and latent factor vectors with respect to the user and the content.

13. A non-transitory machine readable medium having recorded thereon information for adjusting one or more parameters associated with a model, wherein the information, when read by a computer, causes the machine to perform the steps of:

obtaining, from a first source, first information related to activity of a user;

adjusting one or more parameters associated with a model based on the first information obtained within a first time period having a first length of time;

obtaining, from a second source, second information related to activity of the user;

adjusting at least the one or more parameters associated with the model based on the second information obtained within a second time period having a second length of time and a measure indicative of performance of the model;

changing the first length of time when the adjustment of the one or more parameters based on the first information exceeds a first threshold; and

changing the second length of time when the adjustment of the at least one or more parameters based on the second information exceeds a second threshold, wherein

the model is used to determine an affiliation between the user and content, and

the second length of time is larger than the first length of time.

14. The medium of claim 13, wherein the second time period overlaps with the first time period.

15. The medium of claim 13, wherein the step of adjusting one or more parameters associated with a model based on the first information includes:

performing an incremental update of values of the one or more parameters based on the first information.

16. The medium of claim 13, wherein the step of adjusting at least the one or more parameters associated with the model based on the second information includes:

training the model using a collaborative filtering approach based on the second information.

17. The medium of claim 13, wherein the affiliation is based on a score computed based on the model and bias with respect to the user and the content.

18. The medium of claim 13, wherein the affiliation is based on a score computed based on the model and latent factor vectors with respect to the user and the content.