Context-based search using spike waves in spiking neural networks

Patent metadata

Field	Value
Publication number	US-11636318-B2
Application number	US-201716647814-A
Country	US
Kind code	B2
Filing date	Dec 15, 2017
Priority date	Dec 15, 2017
Publication date	Apr 25, 2023
Grant date	Apr 25, 2023

Abstract

Techniques and mechanisms for servicing a search query using a spiking neural network. In an embodiment, a spiking neural network receives an indication of a first context of the search query, wherein a set of nodes of the spiking neural network each correspond to a respective entry of a repository. One or more nodes of the set of nodes are each excited to provide a respective cyclical response based on the first context, wherein a first cyclical response is by a first node. Due at least in part to a coupling of the excited nodes, a perturbance signal, based on a second context of the search query, results in a change of the first resonance response relative to one or more other resonance responses. In another embodiment, data corresponding to the first node is selected, based on the change, as an at least partial result of the search query.

First claim

What is claimed is:

1. A computer device for context-based search with a spiking neural network, the computer device comprising circuitry to:
receive, at a first set of nodes of the spiking neural network, a first one or more spike trains which indicate a first context of a context-based search;
receive, at a second set of nodes of the spiking neural network, a second one or more spike trains from the first set of nodes, wherein each node of the second set of nodes corresponds to a respective entry of a repository, wherein a first node of the second set of nodes corresponds to a first entry of the repository, wherein one or more nodes of the second set of nodes each exhibits a respective cyclical response based on the second one or more spike trains, wherein the one or more nodes of the second set is a subset of all nodes of the second set, and wherein the subset includes the first node;
detect the respective cyclical responses by the one or more nodes of the second set of nodes;
identify the one or more nodes of the second set of nodes based on a detection of the cyclical responses;
selectively send a signal to each of the identified one or more nodes of the second set of nodes, wherein the signal indicates a second context of the context-based search;
receive the signal at the first node of the second set of nodes wherein the signal results in a perturbance of the cyclical response by the first node; and
signal, based on a change to the cyclical response by the first node, a selection of the first entry as a result of the context-based search, wherein the change is based on the perturbance.

2. The computer device of claim 1 , wherein, for each node of the subset, the node is coupled to one or more other nodes of the subset each by a respective inhibitor node.

3. The computer device of claim 1 , wherein the perturbance of the cyclical response by the first node includes at least one of:
an addition of one or more signal spikes to the cyclical response by the first node;
a temporary dampening of the cyclical response by the first node;
a temporary inversion of the cyclical response by the first node; or
a temporary magnification of the cyclical response by the first node.

4. The computer device of claim 1 , wherein the one or more nodes of the second set of nodes further includes one or more other nodes, and wherein the change to the cyclical response by the first node includes a change relative to the respective one or more cyclical responses by the one or more other nodes.

5. The computer device of claim 4 , wherein the respective one or more cyclical responses by the one or more other nodes each decrease in magnitude in response to the perturbance.

6. The computer device of claim 1 , wherein the change to the cyclical response by the first node includes an increase to a magnitude of the cyclical response by the first node.

7. The computer device of claim 1 , further comprising circuitry to identify an index value based on the change to the cyclical response by the first node.

8. The computer device of claim 7 , further comprising circuitry to perform a repository entry lookup based on the index value.

9. At least one non-transitory machine readable medium including instructions that, when executed by a machine, cause the machine to perform operations for context-based search with a spiking neural network, the operations comprising:
receiving, at a first set of nodes of the spiking neural network, a first one or more spike trains which indicate a first context of a context-based search;
receiving, at a second set of nodes of the spiking neural network, a second one or more spike trains from the first set of nodes, wherein each node of the second set of nodes corresponds to a respective entry of a repository, wherein a first node of the second set of nodes corresponds to a first entry of the repository, wherein one or more nodes of the second set of nodes each exhibits a respective cyclical response based on the second one or more spike trains, wherein the one or more nodes of the second set is a subset of all nodes of the second set, and wherein the subset includes the first node;
detecting the respective cyclical responses by the one or more nodes of the second set of nodes;
identifying the one or more nodes of the second set of nodes based on a detection of the cyclical responses;
selectively sending a signal to each of the identified one or more nodes of the second set of nodes, wherein the signal indicates a second context of the context-based search;
receiving the signal at the first node of the second set of nodes wherein the signal results in a perturbance of the cyclical response by the first node; and
signaling, based on a change to the cyclical response by the first node, a selection of the first entry as a result of the context-based search, wherein the change is based on the perturbance.

10. The at least one non-transitory machine readable medium of claim 9 , wherein, for each node of the subset, the node is coupled to one or more other nodes of the subset each by a respective inhibitor node.

11. The at least one non-transitory machine readable medium of claim 9 , wherein the perturbance of the cyclical response by the first node includes at least one of:
an addition of one or more signal spikes to the cyclical response by the first node;
a temporary dampening of the cyclical response by the first node;
a temporary inversion of the cyclical response by the first node; or
a temporary magnification of the cyclical response by the first node.

12. The at least one non-transitory machine readable medium of claim 9 , wherein the one or more nodes of the second set of nodes further includes one or more other nodes, and wherein the change to the cyclical response by the first node includes a change relative to the respective one or more cyclical responses by the one or more other nodes.

13. The at least one non-transitory machine readable medium of claim 12 , wherein the respective one or more cyclical responses by the one or more other nodes each decrease in magnitude in response to the perturbance.

14. The at least one non-transitory machine readable medium of claim 9 , wherein the change to the cyclical response by the first node includes an increase to a magnitude of the cyclical response by the first node.

15. The at least one non-transitory machine readable medium of claim 9 , the operations further comprising identifying an index value based on the change to the cyclical response by the first node.

16. The at least one non-transitory machine readable medium of claim 15 , the operations further comprising performing a repository entry lookup based on the index value.

17. A method for context-based search with a spiking neural network, the method comprising:
receiving, at a first set of nodes of the spiking neural network, a first one or more spike trains which indicate a first context of a context-based search;
receiving, at a second set of nodes of the spiking neural network, a second one or more spike trains from the first set of nodes, wherein each node of the second set of nodes corresponds to a respective entry of a repository, wherein a first node of the second set of nodes corresponds to a first entry of the repository, wherein one or more nodes of the second set of nodes each exhibits a respective cyclical response based on the second one or more spike trains, wherein the one or more nodes of the second set is a subset of all nodes of the second set, and wherein the subset includes the first node;
detecting the respective cyclical resp

Assignees

• Intel Corp

Inventors

• Paul Arnab
• Srinivasa Narayan

Classifications

• G06N3/049 Primary

  Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title