Directed behavior in hierarchical temporal memory based system

• US 8,959,039 B2
• Filed: 01/07/2014
• Issued: 02/17/2015
• Est. Priority Date: 02/10/2006
• Status: Active Grant

• Current Assignee: Numenta, Inc.
• Original Assignee: Numenta, Inc.
• Inventors: Hawkins, Jeffrey C., George, Dileep
• Primary Examiner(s): CHEN, ALAN S
• Application Number:
  US14/149,607
• Publication Number:
  US 20140122394A1
• Time in Patent Office: 406 Days
• Field of Search: None
• US Class Current: 706/12
• 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 hierarchy of computing modules is configured to learn a cause of input data sensed over space and time, and is further configured to determine a cause of novel sensed input data dependent on the learned cause. At least one of the computing modules has a sequence learner module configured to associate sequences of input data received by the computing module to a set of causes previously learned in the hierarchy.

Claims

1. A computer-implemented method of controlling motor behaviors, comprising:

associating first level motor behaviors with effectors at a first level node responsive to receiving input data representing observed motor behaviors of the effectors;

associating second level motor behaviors with the first level motor behaviors at a second level node responsive to receiving the input data, one or more first level motor behaviors forming a second level motor behavior;

the second level node generating second level information representing a second level target behavior based on the association between the second level motor behavior and the first level motor behavior responsive to receiving information indicating intended motor behaviors of the effectors, the second level target behavior corresponding to one of the second level motor behaviors;

the first level node generating first level information representing a first level target behavior based on the association between the first level motor behavior and the effectors responsive to receiving the second level information, the first target behavior corresponding to one of the first level behaviors; and

a control module generating commands to the effectors responsive to receiving the first level information.

2. The method of claim 1, wherein the motor behaviors represented by the input data comprise at least one of visual behavior and acoustic behavior.

3. The method of claim 1, wherein the effectors comprise at least a motor and the control module comprises processing circuitry for operating the motor.

4. The method of claim 1, wherein associating the first level motor behaviors comprises pairing each first level motor behavior with one or more effectors.

5. The method of claim 1, wherein the second level motor behaviors comprises a walking motion in a robot, and the first level motor behaviors associated with the walking motion comprises moving effectors of the robot moves parts corresponding to toe and knee.

6. The method of claim 1, wherein generating the commands to the effectors comprises generating a string of the commands responsive to receiving the first level information.

7. A system for controlling a motor behavior, comprising:

a processor; and

memory coupled to the processor, the memory comprising;

a memory network comprising;

a first level node configured to associate first level motor behaviors with effectors responsive to receiving input data representing observed motor behaviors of the effectors in a learning phase, the first level node configured to generate first level information representing a first level target behavior based on the association between the first level motor behavior and the effectors in a control phase, the first level target behavior corresponding to one of the first level motor behaviors; and

a second level node coupled to the first level node and configured to associate second level motor behaviors with the first level motor behaviors responsive to receiving the input data in the learning phase, the second level node configured to generate second level information representing a second level target behavior based on the association between the second level motor behavior and the first level motor behavior responsive to receiving information of intended motor behaviors of the effectors, the second level node configured to send the second level information to the first level node, one or more first level motor behaviors forming a second level motor behavior, the second level target behavior corresponding to one of the second level motor behaviors; and

a control module coupled to the first level node for sending commands to the effectors responsive to receiving the first level information.

8. The system of claim 7, wherein the control module stores a set of primitive commands for operating the effectors, the control module generating a string of the commands responsive to receiving the first level information.

9. A non-transitory computer-readable storage medium structured to store instructions executable by a processor, the instructions when executed cause the processor to:

associate first level motor behaviors and effectors at a first level node responsive to receiving input data representing observed motor behaviors of the effectors;

associate second level motor behaviors with the first level motor behaviors at a second level node responsive to receiving the input data, one or more first level motor behaviors forming a second level motor behavior;

at the second level node, generate second level information representing a second level target behavior based on the association between the second level motor behavior and the first level motor behavior responsive to receiving information of a target motor behavior, the second level target behavior corresponding to one of the second level motor behaviors;

at the first level node, generate first level information representing a first level target behavior based on the association between the first level motor behavior and the effectors responsive to receiving the second level information, the first target behavior corresponding to one of the first level behaviors; and

at a control module, generate commands to the effectors responsive to receiving the first level information.

10. The computer-readable storage medium of claim 9, wherein the motor behaviors are represented by the input data comprise at least one of visual behavior and acoustic behavior.

11. The computer-readable storage medium of claim 9, wherein the effectors comprise a motor and the control module comprises processing circuitry for operating the motor.

12. The computer-readable storage medium of claim 9, wherein the instructions to generate command comprises generating a string of the commands responsive to receiving the first level information.