Point-to-multipoint communication infrastructure for expert-based knowledge feed-back using learning machines

Patent metadata

Field	Value
Publication number	US-9626628-B2
Application number	US-201414165462-A
Country	US
Kind code	B2
Filing date	Jan 27, 2014
Priority date	Dec 31, 2013
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

In one embodiment, techniques are shown and described relating to a point-to-multipoint communication infrastructure for expert-based knowledge feed-back using learning machines. A learning machine may communicate an expert discovery request into a network to discover one or more experts, and then receive from the one or more experts, one or more expert discovery responses. Based on the one or more received expert discovery responses, the learning machine may then build a dynamic multicast tree of experts to assist the learning machine in a computer network.

First claim

What is claimed is:

1. A method, comprising:
communicating, by a node in a network executing a learning machine (LM), an expert discovery request into the network to discover one or more experts;
receiving, from the one or more experts, one or more expert discovery responses;
building, based on the one or more expert discovery responses, a dynamic multicast tree of experts to create a communication infrastructure for the LM to interact with multiple experts for supervised learning;
determining, by the node, that the LM does not have a way of categorizing a certain set of data received at the node;
in response to determining that the LM cannot categorize the certain set of data, triggering, by the node, an on-the-fly tunnel connection to at least one of the one or more experts in the multicast tree of experts;
sending, by the node, an expert request message over the on-the-fly tunnel connection, the expert request message requesting input by the at least one expert to categorize the certain set of data for the LM;
receiving, at the node, one or more expert input responses; and
categorizing, by the LM on the node, the certain set of data based on the one or more expert input responses.

2. The method according to claim 1 , further comprising:
including in the expert request message an indication of a particular expert from which the LM is requesting information.

3. The method according to claim 1 , further comprising:
categorizing, based on the one or more expert discovery responses, the experts into one or more areas of expertise; and
dividing, based on the one or more areas of expertise, the dynamic multicast tree of experts into one or more dynamic multicast sub-trees of experts, wherein an expert can be associated with more than one dynamic multicast sub-tree of experts simultaneously.

4. The method according to claim 1 , wherein the one or more expert discovery responses include data on one or more expert parameters selected from the group consisting of: area of expertise, availability, inquiry response time, response accuracy, and severity of issues.

5. The method according to claim 1 , further comprising:
evaluating, based on the received one or more expert discovery responses, the one or more experts.

6. The method according to claim 5 , wherein the evaluating is based on one or more parameters, wherein the one or more parameters include a quality of feedback, a response time for providing feedback, an accuracy of feedback, an expertise associate with a particular topic, an average expertise across many topics, a cost associated with feedback, a subsequent improvement in network performance, or a review from another expert.

7. The method according to claim 5 , further comprising:
recommending, based on the evaluating, a particular expert for a particular LM inquiry.

8. The method according to claim 1 , further comprising:
dynamically monitoring a status of the one or more experts; and
de-registering any of the one or more experts that are no longer available to assist the LM, wherein the de-registering is initiated by the LM or the one or more experts.

9. The method according to claim 1 , further comprising:
maintaining a dedicated communication channel for the one or more experts to register and join.

10. An apparatus, comprising:
one or more network interfaces that communicate with a network;
a processor coupled to the one or more network interfaces and configured to execute a process; and
a memory configured to store program instructions which contain the process executable by the processor, the process comprising:
communicating an expert discovery request into a network to discover one or more experts;
receiving, from the one or more experts, one or more expert discovery responses;
building, based on the one or more expert discovery responses, a dynamic multicast tree of experts to create a communication infrastructure for the LM to interact with multiple experts for supervised learning;
determining that the LM does not have a way of categorizing a certain set of data received at the node;
in response to determining that the LM cannot categorize the certain set of data, triggering an on-the-fly tunnel connection to at least one of the one or more experts in the multicast tree of experts;
sending an expert request message over the on-the-fly tunnel connection, the expert request message requesting input by the at least one expert to categorize the certain set of data for the LM;
receiving one or more expert input responses; and
categorizing, by the LM, the certain set of data based on the one or more expert input responses.

11. The apparatus according to claim 10 , wherein the process further comprises:
including in the expert request message an indication of a particular expert from which the LM is requesting information.

12. The apparatus according to claim 10 , wherein the process further comprises:
categorizing, based on the one or more expert discovery responses, the experts into one or more areas of expertise; and
dividing, based on the one or more areas of expertise, the dynamic multicast tree of experts into one or more dynamic multicast sub-trees of experts, wherein an expert can be associated with more than one dynamic multicast sub-tree of experts simultaneously.

13. The apparatus according to claim 10 , wherein the one or more expert discovery responses include data on one or more expert parameters selected from the group consisting of: area of expertise, availability, inquiry response time, response accuracy, and severity of issues.

14. The apparatus according to claim 10 , wherein the process further comprises:
evaluating, based on the received one or more expert discovery responses, the one or more experts.

15. The apparatus according to claim 10 , wherein the evaluating is based on one or more parameters, wherein the one or more parameters include a quality of feedback, a response time for providing feedback, an accuracy of feedback, an expertise associate with a particular topic, an average expertise across many topics, a cost associated with feedback, a subsequent improvement in network performance, or a review from another expert.

16. The apparatus according to claim 10 , wherein the process further comprises:
recommending, based on the evaluating, a particular expert for a particular LM inquiry.

17. The apparatus according to claim 10 , wherein the process further comprises:
dynamically monitoring a status of the one or more experts; and
de-registering any of the one or more experts that are no longer available to assist the LM, wherein the de-registering is initiated by the LM or the one or more experts.

18. A tangible non-transitory computer readable medium storing program instructions that cause a computer to execute a process, the process comprising:
communicating, from a node executing a learning machine (LM), an expert discovery request to one or more experts;
receiving, from the one or more experts, one or more expert discovery responses;
building, based on the one or more expert discovery responses, a dynamic multicast tree of experts to create a communication infrastructure for the LM to interact with multiple experts for supervised learning;
determining, as the node, that the LM does not have a way of categorizing a certain set of data received at the node;
in response to determining that the LM cannot categorize the certain set of data, triggering, at the node, an on-the-fly tunnel connection to at least one of the one or more experts in the multicast tree of

Assignees

• Cisco Tech Inc

Inventors

• Dasgupta Sukrit
• Vasseur Jean-Philippe
• Mermoud Grégory

Classifications

• G06N99/005 Primary

  Physics · mapped topic

• H04L41/16 Primary

  using machine learning or artificial intelligence · CPC title

• G06N20/00 Primary

  Machine learning · CPC title