Hierarchical temporal memory (HTM) system deployed as web service

• US 10,516,763 B2
• Filed: 03/03/2017
• Issued: 12/24/2019
• Est. Priority Date: 02/10/2006
• Status: Active Grant

• Current Assignee: Numenta, Inc.
• Original Assignee: Numenta, Inc.
• Inventors: Edwards, Jeffrey L., Saphir, William C., Ahmad, Subutai, George, Dileep, Astier, Frank, Marianetti, Ronald
• Primary Examiner(s): Starks, Wilbert L
• Application Number:
  US15/449,753
• Publication Number:
  US 20170180515A1
• Time in Patent Office: 1,026 Days
• Field of Search: None
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06V 10/95 structured as a network, e.g. client-server architectures
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 67/01 Protocols
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: H04L 67/1017 based on a round robin mechanism
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

A web-based hierarchical temporal memory (HTM) system in which one or more client devices communicate with a remote server via a communication network. The remote server includes at least a HTM server for implementing a hierarchical temporal memory (HTM). The client devices generate input data including patterns and sequences, and send the input data to the remote server for processing. The remote server (specifically, the HTM server) performs processing in order to determine the causes of the input data, and sends the results of this processing to the client devices. The client devices need not have processing and/or storage capability for running the HTM but may nevertheless take advantage of the HTM by submitting a request to the HTM server.

Claims

1. A method of controlling one or more effectors, comprising:

causing movements of one or more physical effectors by invoking one or more predefined behaviors stored in a control module;

receiving from a sensor, by a learning module, first sensed inputs resulting from the movements of the one or more physical effectors that occur responsive to invoking the one or more behaviors at the control module;

generating, by the learning module, a representation of relationships between the invoked predefined one or more behaviors and the movements of the one or more physical effectors based on at least the first sensed inputs and the one or more invoked behaviors; and

making prediction based on the representation of the relationships to cause an intended behavior by controlling the one or more physical effectors.

2. The method of claim 1, further comprising:

associating each of first level nodes in a plurality of nodes in the learning module with a first level effector motion; and

associating each of second level nodes in the plurality of nodes with a second level effector motion, one or more first level effector motions combined to form a second level effector motion.

3. The method of claim 2, further comprising switching off a node of the plurality of nodes to prevent first or second level effector motions corresponding to the node.

4. The method of claim 3, further comprising modifying a signal from a node of the plurality of nodes to change second level effector motions corresponding to the node.

5. The method of claim 4, wherein at least a subset of the one or more effectors controls capturing of images, and the first sensed inputs represent the captured images.

6. The method of claim 5, wherein the one or more effectors comprise at least a motor and the control module comprises processing circuitry for operating the motor.

7. The method of claim 1, where the relationships pair at least a predefined behavior with at least a subset of the effectors that causes the predefined behavior.

8. The method of claim 1, wherein the learning modules comprise Hierarchical Temporal Memory module.

9. The method of claim 1, further comprising:

receiving from a source, by the learning module, second sensed inputs irrespective of the movements of the one or more effectors, the representation of relationships based further on the second sensed inputs; and

predicting patterns or sequences to appear in third sensed input from the source based on the representation of the relationship.

10. A computer-implemented method, comprising:

receiving first input data by a first node of a first level of hierarchy;

receiving, by the first node, a control signal from a second node of a second level of hierarchy higher than the first level, the control signal assigning a first spatial pattern and a second spatial pattern to a same cause;

storing, by the first node, relationships of spatial patterns and temporal sequences based on the first input data and the control signal from the second node;

generating, by the first node, information indicating detecting of the first spatial pattern responsive to detecting the first spatial pattern in second input data; and

generating, by the first node, the same information indicating detecting of the first spatial pattern responsive to receiving a second pattern in the second input data.

11. The computer-implemented method of claim 10, further comprising:

generating, by the first node, the same information indicating detecting of the first pattern responsive to detecting a third spatial pattern in the second input data within a spatial distance from the first spatial pattern.

12. The computer-implemented method of claim 10, wherein the second spatial pattern is not within the spatial distance from the first spatial pattern.

13. The computer-implemented method of claim 10, further comprising outputting, from the first node to the second node, a node output generated based on the coincidence information.

14. The computer-implemented method of claim 13, further comprising outputting, from a third node to the second node, another node output, the third node at the first level of hierarchy.

15. The computer-implemented method of claim 10, wherein the first input data is received by the first node during a learning stage, and the second input data is received by the first node during an inference stage subsequent to the learning stage.

16. The computer-implemented method of claim 10, wherein the second spatial pattern is determined to be within the spatial distance from the first spatial pattern by comparing bits in the first spatial pattern with bits in the second spatial pattern.

17. The computer-implemented method of claim 10, further comprising:

causing movements of one or more effectors by invoking one or more predefined behaviors stored in a control module; and

generating, by a sensor, the first input data resulting from the movements of the one or more effectors that occur responsive the movement of one or more effectors.

18. A computer-implemented method, comprising:

receiving first input data by a first node of a first level of hierarchy;

receiving, by the first node, a control signal from a second node of a second level of hierarchy higher than the first level, the control signal assigning a first spatial pattern and a second spatial pattern to a same cause;

pooling, by the first node, spatial patterns in the first input data into one or more temporal sequences based on sequential appearance of the spatial patterns in the first input data and the control signal from the second node, the first spatial pattern and the second spatial pattern assigned to a same temporal sequence; and

outputting, by the first node to the second node, a node output representing detected sequences of patterns in a second input data based on the pooling.

19. The computer-implemented method of claim 18, wherein the second spatial pattern is not within a predetermined temporal proximity from the first spatial pattern in the first input data.

20. The computer-implemented method of claim 18, further comprising:

generating, by the first node, coincident information responsive to detecting the first pattern in the second input data;

generating, by the first node, the same coincident information responsive to receiving a second spatial pattern in the second input data; and

generating, by the first node, the node output based on the coincident information.