SEQUENCE LEARNING IN A HIERARCHICAL TEMPORAL MEMORY BASED SYSTEM

• US 20080059389A1
• Filed: 01/11/2007
• Published: 03/06/2008
• Est. Priority Date: 02/10/2006
• Status: Active Grant

• Current Assignee: Numenta, Inc.
• Original Assignee: Numenta, Inc.
• Inventors: Jaros, Robert, Hawkins, Jeffrey, George, Dileep, Astier, Frank
• Granted Patent: US 7,620,608 B2
• Time in Patent Office: Days
• 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 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, wherein at least one of the computing modules comprises;

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.

2. The system of claim 1, wherein the computing module further comprises:

a coincidence detection module operatively connected to the sequence learner module and configured to detect coincidences among input data received by the computing module,

wherein the sequences of input data received by the sequence learner module are provided by the coincidence detection module.

3. The system of claim 1, wherein the sequence learner module is further configured to output a distribution representing probabilities of possible causes of the input data received by the computing module, the distribution being dependent on the association of sequences to the set of previously learned causes.

4. The system of claim 3, wherein the sequence learner has a predetermined number of outputs.

5. The system of claim 3, wherein the sequence learner module is further configured to adjust the distribution as input data to the computing module changes over time.

6. The system of claim 1, wherein the sequence learner module is further configured to store a count of how many times a particular sequence has been received by the computing module.

7. The system of claim 6, wherein the sequence learner module is further configured to store the count dependent on whether the particular sequence has been received by the computing module at a statistically frequent rate.

8. The system of claim 1, wherein the sequence learner module is further configured to associate a sequence to the set of previously learned causes dependent on the sequence being identified as being received by the computing module at a statistically frequent rate.

9. The system of claim 1, wherein the sequence learner module is further configured to collect data on received sequences that are shorter than a predetermined sequence length.

10. The system of claim 1, wherein the sequence learner module is further configured to collect data on received sequences that have been identified as being statistically frequent based on observations of shorter length sequences.

11. The system of claim 1, wherein the sequence learner module is further configured to remove a sequence from being represented at an output of the sequence learner module.

12. A computer-implemented method, comprising:

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

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

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

13. The computer-implemented method of claim 12, further comprising:

outputting the distribution over a fixed number of outputs.

14. The computer-implemented method of claim 12, further comprising:

adjusting the distribution as inputted spatial patterns change over time.

15. The computer-implemented method of claim 12, further comprising:

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

16. The computer-implemented method of claim 15, further comprising:

storing the count dependent on whether the particular sequence has been received at a statistically frequent rate.

17. The computer-implemented method of claim 12, further comprising:

associating a received sequence to the set of previously learned causes dependent on the sequence being identified as being received at a statistically frequent rate.

18. The computer-implemented method of claim 12, further comprising:

collecting data on received sequences that are shorter than a predetermined sequence length.

19. The computer-implemented method of claim 12, further comprising:

collecting data on received sequences that have been identified as being statistically frequent based on observations of shorter length sequences.

20. The computer-implemented method of claim 12, further comprising:

removing a sequence from being represented in the outputted distribution.

21. A 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 probabilities of the statistically frequent sequences being a cause of the sensed input data, wherein the distribution is generated over a set of previously learned causes.

22. The computer-readable medium of claim 21, further comprising instructions to:

output the distribution over a fixed number of outputs.

23. The computer-readable medium of claim 21, further comprising instructions to:

adjust the distribution as inputted spatial patterns change over time.

24. The computer-readable medium of claim 21, further comprising instructions to:

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

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

store the count dependent on whether the particular sequence has been received at a statistically frequent rate.

26. The computer-readable medium of claim 21, further comprising instructions to:

associate a received sequence to the set of previously learned causes dependent on the sequence being identified as being received at a statistically frequent rate.

27. The computer-readable medium of claim 21, further comprising instructions to:

collect data on received sequences that are shorter than a predetermined sequence length.

28. The computer-readable medium of claim 21, further comprising instructions to:

collect data on received sequences that have been identified as being statistically frequent based on observations of shorter length sequences.

29. The computer-readable medium of claim 21, further comprising instructions to:

remove a sequence from being represented in the outputted distribution.