NEURAL NETWORK SYSTEM AND METHOD FOR CONTROLLING OUTPUT BASED ON USER FEEDBACK

• US 20130018833A1
• Filed: 07/12/2012
• Published: 01/17/2013
• Est. Priority Date: 02/02/1999
• Status: Abandoned Application

• Current Assignee: Tamiras Per Pte Ltd LLC (Intellectual Ventures LLC)
• Original Assignee: Garbortese Holdings LLC (Intellectual Ventures LLC)
• Inventors: Pope, Ivan, Sullivan, Alan T
• Application Number:
  US13/547,778
• Publication Number:
  US 20130018833A1
• Time in Patent Office: Days
• US Class Current: 706/25
• CPC Class Codes: G06F 16/3326 using relevance feedback from the user, e.g. relevance feedback on documents, documents sets, document terms or passages
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/063 using electronic means
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 69/329 in the application layer [OSI layer 7]

Abstract

For various information sources, information output based on user feedback about information from the sources is controlled. A neural network module selects object(s) to receive information from the information sources based on inputs and weight values during that epoch. A server, associated with the neural network module, provides the object(s) to recipients. The object(s) may comprise electronic mail messages, chat participants viewers, or slots within a link directory page. The recipients provide feedback about the information during an epoch. At the conclusion of an epoch, the neural network takes the feedback provided by the recipients and generates a rating value for the object(s). Based on the rating value and the selections made, the neural network re-determines the weight values within the network. The neural network then selects the object(s) to receive information during a subsequent epoch using the re-determined weight values and the inputs for that subsequent epoch.

Claims

1. A system, comprising:

a processor; and

a memory communicatively coupled to the processor and having stored thereon computer-executable components, including;

a node evaluation module configured to determine, based on weight values and threshold values for respective nodes of a neural network, output values for respective output nodes of the neural network given a set of input values of respective input nodes of the neural network; and

a learning computation module configured to modify the weight values at an end of an epoch based on feedback regarding the output values received during the epoch.

2. The system of claim 1, further comprising an application module configured to select a first set of links to web pages for presentation at a client device based on the output values given the set of input values, wherein the set of input values represent respective attributes of a first web page provided to the respective input nodes in response to a visit to the first web page by the client device.

3. The system of claim 2, wherein the application module is further configured to receive, as the feedback, a selection received by the client device of at least one link in the first set of links during the epoch, wherein the epoch comprises a duration of a visit to the first web page by the client device.

4. The system of claim 2, wherein the application module is further configured to receive, as the feedback, relevance input indicating a relevance of at least one of the first set of links.

5. The system of claim 2, wherein the neural network module is further configured to learn, based on the one or more weight values, that navigation to the first web page is likely to be followed by selection of a subset of links of the first set of links.

6. The system of claim 5, wherein, in response to a subsequent visit to the first web page by the client, the application module is further configured to select a second set of links for presentation at the client device based on second output values for the respective output nodes of the neural network, wherein the second output values are based on the set of input values and the weight values as modified by the learning computation module.

7. The system of claim 6, wherein the neural network module is further configured to establish an association between the feedback and one or more keywords in a query used to invoke the first web page, and determine whether to include the subset of links in the second set of links based on the association in response to a subsequent query including the one or more keywords.

8. A method, comprising:

determining, by a system including a processor, output values for respective output nodes of a neural network based on weight values and threshold values for respective nodes of the neural network and a set of input values of respective input nodes of the neural network; and

modifying at least one of the weight values after an end of an epoch based on feedback regarding the output values received during the epoch to yield at least one modified weight value.

9. The method of claim 8, further comprising selecting a first plurality of links to network locations for delivery to a client device based on the output values determined based on the set of input values, wherein the set of input values represent a respective set of attributes of a first network location input to the respective input nodes in response to the client device browsing to the first network location.

10. The method of claim 9, further comprising receiving, as the feedback during the epoch, a selection by the client device of at least one link in the first plurality of links, wherein the epoch comprises a duration of time that the client device visits the first network location.

11. The method of claim 9, further comprising receiving, as the feedback, relevance feedback quantifying a relevance of at least one of the first plurality of links.

12. The method of claim 9, further comprising determining, based on the one or more weight values, that navigation to the first network location has a probability of being followed by selection of a subset of links of the first plurality of links.

13. The method of claim 12, further comprising selecting, in response to a subsequent visit to the first network location by the client device, a second plurality of links for presentation to the client device based on second output values for the respective output nodes of the neural network, wherein the second output values are based on the set of input values and the at least one modified weight value.

14. The method of claim 13, further comprising:

creating an association between the feedback and one or more keywords in a query used to invoke the first network location, and

determining whether to include the subset of links in the second plurality of links based on the association in response to a subsequent query including the one or more keywords.

15. A non-transitory computer-readable medium having stored thereon computer-executable instructions that, in response to execution, cause a computing device to perform operations, comprising:

calculating, based on weight values and threshold values for respective nodes of a neural network, first output values for respective output nodes of the neural network given a set of input values provided to respective input nodes of the neural network; and

in response to determining an epoch has ended, modifying at least one of the weight values at an end of an epoch to generate at least one modified weight value, wherein the modifying is based on feedback relating to the output values received during the epoch.

16. The non-transitory computer-readable medium of claim 15, wherein the operations further comprise selecting a first plurality of links to web sites to be rendered on a client device based on the output values given the set of input values, wherein the set of input values represent a respective set of attributes of a first web site provided to the respective input nodes in response to the client device visiting the first web site.

17. The non-transitory computer-readable medium of claim 16, wherein the operations further comprise receiving, as the feedback, a selection by the client device of at least one link in the first plurality of links during the epoch, wherein the epoch comprises a duration that the client device visits the first web site.

18. The non-transitory computer-readable medium of claim 16, wherein the operations further comprise learning that navigation to the first web site has a likelihood of being followed by selection of a subset of links of the first plurality of links based on the one or more weight values.

19. The non-transitory computer-readable medium of claim 18, wherein the operations further comprise selecting, in response to the client device visiting the first web site after the end of the epoch, a second plurality of links to be rendered on the client device based on second output values for the respective output nodes of the neural network, wherein the second output values are based on the set of input values and the at least one modified weight value.

20. The transitory computer-readable medium of claim 19, wherein the operations further comprise:

associating the feedback with one or more keywords of a query used to invoke the first web site, and

including the subset of links in the second plurality of links based on the associating in response to a subsequent query including the one or more keywords.