Architecture of a hierarchical temporal memory based system

• US 8,447,711 B2
• Filed: 04/14/2008
• Issued: 05/21/2013
• Est. Priority Date: 02/10/2006
• Status: Active Grant

• Current Assignee: Numenta, Inc.
• Original Assignee: Numenta, Inc.
• Inventors: Hawkins, Jeffrey C., Ahmad, Subutai, George, Dileep, Astier, Frank E., Marianetti, Ronald II
• Primary Examiner(s): CHEN, ALAN S
• Application Number:
  US12/052,580
• Publication Number:
  US 20080183647A1
• Time in Patent Office: 1,863 Days
• Field of Search: None
• US Class Current: 706/15
• CPC Class Codes: G06N 20/00 Machine learning
• 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: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A hierarchical temporal memory (HTM) based system may be provided as a software platform. The software platform includes: a runtime engine arranged to run an HTM network; a first interface accessible by a set of tools to configure, design, modify, train, debug, and/or deploy the HTM network; and a second interface accessible to extend a functionality of the runtime engine.

Claims

1. A system, comprising:

an Hierarchical Temporal Memory (HTM) network executable at least in part on a central processor unit (CPU), the HTM network comprising;

a first node configured to receive an input data and generate a first vector representing information about corresponding of patterns and sequences in the input data to learned patterns and sequences; and

a second node configured to receive the first vector and generate a second vector based on the first vector, the second vector representing information about causes of the input data; and

a supervisor entity configured to;

manage communication between a user application and the HTM network; and

configure connective relationships of nodes in the HTM network responsive to receiving configuration information for the HTM network.

2. The system of claim 1, wherein the supervisor entity is instantiated at least in part in software.

3. The system of claim 1, wherein the HTM network is implemented on a server including one or more central processing units (CPUs).

4. The system of claim 1, wherein the communication comprises an XML-encoded message.

5. The system of claim 1, wherein the communication is transmitted using at least one of an HTTP and an HTTPS protocol.

6. The system of claim 1, wherein the HTM network is executable across a plurality of CPUs.

7. A software platform, comprising:

an Hierarchical Temporal Memory (HTM) network comprising;

a first node configured to receive an input data and generate a first vector representing information about corresponding of patterns and sequences in the input data to learned patterns and sequences; and

a second node configured to receive the first vector and generate a second vector based on the first vector, the second vector representing information about causes of the input data; and

a supervisor entity configured to;

manage communication between a user application and the HTM network; and

configure connective relationships of nodes in the HTM network responsive to receiving configuration information for the HTM network.

8. The software platform of claim 7, wherein the HTM network is distributed over a plurality of central processing units.

9. The software platform of claim 7, wherein the HTM network is operated with a single central processing unit.

10. The software platform of claim 7 further comprising:

a node processing unit for managing at least a portion of the HTM network.

11. The software platform of claim 10, wherein the node processing unit is allocated to a CPU running the at least a portion of the HTM network to manage a portion of the HTM network running on the CPU.

12. The software platform of claim 7, wherein the supervisor entity comprises a net list for describing a configuration of the HTM network.

13. The software platform of claim 7, wherein the supervisor entity is arranged to allocate the at least a portion of the HTM network to a central processing unit.

14. The software platform of claim 7 further comprising a node plug-in interface for dynamically extending functionality of the HTM network.

15. The software platform of claim 7 further comprising a configuration tool associated with the supervisor entity for configuring the HTM network, and a training tool associated with the supervisor entity for training the HTM network.

16. A method of operating an Hierarchical Temporal Memory (HTM) network, comprising:

creating the HTM network at least partly on a first computer system, via a supervisor entity provided on the first computer system, the HTM network comprising a first node and a second node, the first node configured to receive an input data and generate a first vector representing information about patterns and sequences in the input data corresponding to learned patterns and sequences, the second node configured to generate a second vector based on the first vector, the second vector representing information about causes of the input data;

modifying connective relationships of nodes in the created HTM network via the supervisor entity;

training the created or modified HTM network via the supervisor entity; and

outputting the second vector to a user application via the supervisor entity.

17. The method of claim 16, wherein the first computer system comprises a plurality of central processing units.

18. The method of claim 16, further comprising:

extending a functionality of the HTM network by providing a node plug-in interface for associating a third node with the first node or the second node, the third node provided on a second computing system.

19. The method of claim 16 further comprising:

communicating a message between a user application and the supervisor entity.

20. The method of claim 16, further comprising:

running the HTM network responsive to instructions provided to the computer system via the supervisor entity.

21. A computer-implemented method, comprising:

implementing a network of a hierarchy of computing modules using a subclass extending from a base class defined using object-oriented programming;

implementing a supervisor entity for managing communication between the network and a user application based on another subclass extending from the base class;

running the network to train the network with a cause of a first set of input data received over space and time; and

determine by the network a cause of a second set of input data dependent on the trained cause.

22. The method of claim 21, further comprising:

dynamically linking a library during operation of the network, wherein the library comprises code for extending the base class.

23. A non-transitory computer-readable medium having instructions stored therein that are executable on a processor, the instructions comprising instructions to:

input spatial patterns in sensed input data, wherein spatial patterns received over time represent sequences;

identify received sequences that occur frequently according to a predetermined statistical threshold; and

output a distribution representing information about the statistically frequent sequences being a cause of the sensed input data, wherein the distribution is generated over a set of previously learned causes.

24. The computer-readable medium of claim 23, further comprising instructions to:

store a count of how many times a particular sequence has been received.

25. A system, comprising:

a processor, and

a hierarchy of computing modules configured to learn a cause of input data sensed over space and time, the hierarchy further configured to determine a cause of novel sensed input data dependent on the learned cause, and the hierarchy having a first level of computing modules and a second level of at least one computing module,

wherein at least one of the computing modules in the first level is configured to receive a portion of the novel sensed input data,

wherein the computing module in the first level is further capable of determining a possible cause of the novel sensed input data dependent on analyzing only a subset of the portion of the novel sensed input data, and

wherein the subset is determined dependent on a control signal received by the computing module in the first level.

26. The system of claim 25, wherein the control signal is generated by the computing module in the second level, wherein an input range of the computing module in the second level is greater than an input range of the computing module in the first level.

27. A system, comprising:

a hierarchy of computing modules configured to learn a cause of input data sensed over space and time, the hierarchy further configured to determine a cause of novel sensed input data dependent on the learned cause, and the hierarchy having a first level of computing modules and a second level of at least one computing module, wherein at least one of the computing modules in the first level operates on a first server, and wherein the at least one computing module in the second level operates on a second server; and

at least one message manager module configured to relay information between the first server and the second server, wherein the message manager module is further configured to operate according to at least one of a message passing interface (MPI) protocol and a zero-copy protocol using shared memory.

28. A system, comprising:

a hierarchy of computing modules configured to learn a cause of input data sensed over space and time, the hierarchy further configured to determine a cause of novel sensed input data dependent on the learned cause, and the hierarchy having a first level of computing modules and a second level of at least one computing module, wherein at least one of the computing modules in the first level operates on a first server, and wherein the at least one computing module in the second level operates on a second server; and

at least one message manager module configured to relay information between the first server and the second server, wherein the message manager module is further configured to operate dependent on at least one of a socket connection and a shared memory buffer.