Regulating activation threshold levels in a simulated neural circuit

1. Computer hardware configured to implement a simulated neural element comprising:

at least two branches, each branch being configured to:

receive one or more input signals in response to sensory event occurrences, and

activate in response to a combination of the input signals received during a first specified window of time exceeding a branch activation threshold level required to activate the branch; and

a simulated cell body configured to activate in response to a combination of activity in the branches during a second specified window of time exceeding a cell body activation threshold level, wherein the simulated neural element is configured to:

regulate the branch activation threshold levels such that some measures of the actual branch activation rates for the branches approximate desired branch activation rate measures; and

regulate the cell body activation threshold level such that some measure of the actual cell body activation rate for the simulated cell body approximates a desired cell body activation rate measure.

2. The computer hardware implementing the simulated neural element of claim 1 wherein:

the measure of the actual branch activation rate for a branch is a measure of branch activation frequency relative to a total number of sensory event occurrences during some preceding time period; and

the measure of the actual cell body activation rate is a measure of cell body activation frequency relative to a total number of sensory event occurrences during some preceding time period.

3. The computer hardware implementing the simulated neural element of claim 2 wherein the desired branch activation rate measures are the same for all of the branches.

4. The computer hardware implementing the simulated neural element of claim 2 wherein the desired branch activation rate measures vary for the branches.

5. The computer hardware implementing the simulated neural element of claim 2 wherein each branch is coupled to a plurality of inputs and is configured to:

receive input signals from the inputs to which the branch is coupled; and

activate when some combination of concurrently received input signals from the inputs to which the branch is coupled exceeds the branch's branch activation threshold level.

6. The computer hardware implementing the simulated neural element of claim 5 wherein:

the input signals received from the inputs are binary; and

each branch is configured to activate when a weighted sum of the binary input signals received, during the first specified window of time, from the inputs to which the branch is coupled exceeds the branch's branch activation threshold level.

7. The computer hardware implementing the simulated neural element of claim 5 wherein:

the input signals received from the inputs reflect levels of activity; and

each branch is configured to activate when a sum of the input signals received, during the first specified window of time, from the inputs to which the branch is coupled exceeds the branch's branch activation threshold level.

8. The computer hardware implementing the simulated neural element of claim 7 wherein the sum of the input signals comprises a weighted sum of the input signals received, during the first specified window of time, from the inputs to which the branch is coupled.

9. The computer hardware implementing the simulated neural element of claim 7 wherein the sum of the input signals comprises a linear sum of the input signals received, during the first specified window of time, from the inputs to which the branch is coupled.

10. The computer hardware implementing the simulated neural element of claim 5 wherein each branch is configured to activate when a non-linear combination of the input signals received, during the first specified window of time, from the inputs to which the branch is coupled exceeds the branch's branch activation threshold level.

11. The computer hardware implementing the simulated neural element of claim 2 wherein:

the cell body activation threshold level is a threshold number of activated branches required to activate the cell body; and

the cell body is configured to activate in response to a number of activated branches of the simulated neural element during the second specified window of time exceeding the threshold number of activated branches required to activate the cell body.

12. The computer hardware implementing the simulated neural element of claim 2 wherein the simulated neural element is configured to:

increase a branch's branch activation threshold level when the measure of the branch's actual branch activation rate is greater than the branch's desired branch activation rate; and

decrease the branch's branch activation threshold level when the measure of the branch's actual branch activation rate is less than the branch's desired branch activation rate.

13. The computer hardware implementing the simulated neural element of claim 2 wherein the simulated neural element is configured to:

increase the cell body activation threshold level when the measure of the actual cell body activation rate is greater than the desired cell body activation rate; and

decrease the cell body activation threshold level when the measure of the actual cell body activation rate is less than the desired branch activation rate.

14. The computer hardware implementing the simulated neural element of claim 13 wherein:

the cell body activation threshold level is a threshold number of activated branches required to activate the cell body;

the simulated cell body is configured to activate in response to a number of activated branches of the simulated neural element during the second specified window of time exceeding the threshold number of activated branches required to activate the cell body; and

the simulated neural element is configured to:

increase the threshold number of activated branches required to activate the cell body when the measure of the actual cell body activation rate is greater than the desired cell body activation rate; and

decrease the threshold number of activated branches required to activate the cell body when the measure of the actual cell body activation rate is less than the desired branch activation rate.

15. The computer hardware implementing the simulated neural element of claim 1 wherein:

each branch is configured to regulate the branch's branch activation threshold level such that the measure of the actual branch activation rate for the branch approximate the branch's desired branch activation rate measure; and

the simulated cell body is configured to regulate the cell body activation threshold level such that the measure of the actual cell body activation rate for the simulated cell body approximates the desired cell body activation rate measure.

16. A method for regulating activity levels in a simulated branched neural element of a simulated neural circuit, the simulated branched neural element including at least two branches and a simulated cell body, each branch being configured to receive indications of input activity levels and to activate in response to combinations of the input activity levels exceeding a branch activation threshold level, the simulated cell body being configured to activate in response to combinations of activity in the branches exceeding a cell body activation threshold level, the method comprising:

establishing, using at least one computer, desired activity rates for the branches;

establishing, using the at least one computer, a desired activity rate for the simulated cell body;

monitoring, using the at least one computer, actual activity rates of the branches;

monitoring, using the at least one computer, an actual activity rate of the simulated cell body;

regulating, using the at least one computer, the branch activation threshold levels such that the actual activity rates of the branches approximate the desired activity rates for the branches; and

regulating, using the at least one computer, the cell body activation threshold level such that the actual activity rate of the simulated cell body approximates the desired activity rate for the simulated cell body.

17. The method of claim 16 wherein:

monitoring actual activity rates of the branches comprises averaging activity in the branches during a window in time; and

monitoring an actual activity rate of the simulated cell body comprises averaging activity in the simulated cell body during a window in time.

18. The method of claim 16 wherein:

averaging activity in the branches during a window in time includes weighting activity based on its recency relative to the window in time; and

averaging activity in the simulated cell body during a window in time includes weighting activity based on its recency relative to the window in time.

19. The method of claim 16 wherein:

the cell body activation threshold level comprises a number of activated branches required to activate the simulated cell body; and

the simulated cell body is configured to activate in response to a number of the simulated branched neural element's branches that are activated during a specified window of time exceeding the number of activated branches required to activate the simulated cell body.

20. The method of claim 16 wherein regulating a branch's branch activation threshold level comprises:

increasing the branch's branch activation threshold level when the branch's actual activity rate is greater than the branch's desired activity rate; and

decreasing the branch's branch activation threshold level when the branch's actual activity rate is less than the branch's desired activity rate.

21. The method of claim 16 wherein regulating the cell body activation threshold level comprises:

increasing the simulated cell body's activation threshold level when the simulated cell body's actual activity rate is greater than the desired activity rate for the simulated cell body; and

decreasing the simulated cell body's activation threshold level when the simulated cell body's actual activity rate is less than the desired activity rate for the simulated cell body.

22. The method of claim 21 wherein:

the cell body activation threshold level comprises a number of activated branches required to activate the simulated cell body;

the simulated cell body is configured to activate in response to a number of the simulated branched neural element's branches that are activated during a specified window of time exceeding the number of activated branches required to activate the simulated cell body; and

regulating the cell body activation threshold level comprises:

increasing the number of activated branches required to activate the simulated cell body when the simulated cell body's actual activity rate is greater than the desired activity rate for the simulated cell body, and

decreasing the number of activated branches required to activate the simulated cell body when the simulated cell body's actual activity rate is less than the desired activity rate for the simulated cell body.

23. A computer program for regulating activity levels in a simulated branched neural element of a simulated neural circuit, the simulated branched neural element including at least two branches and a simulated cell body, each branch being configured to receive indications of input activity levels and to activate in response to combinations of the input activity levels exceeding a branch activation threshold level, the simulated cell body being configured to activate in response to combinations of activity in the branches exceeding a cell body activation threshold level, the computer program being embodied on a tangible computer-readable medium and including instructions that, when executed, cause a computer to:

establish desired activity rates for the branches;

establish a desired activity rate for the simulated cell body;

monitor actual activity rates of the branches;

monitor an actual activity rate of the simulated cell body;

regulate the branch activation threshold levels such that the actual activity rates of the branches approximate the desired activity rates for the branches; and

regulate the cell body activation threshold level such that the actual activity rate of the simulated cell body approximates the desired activity rate for the simulated cell body.

24. The computer program of claim 23 wherein:

the instructions that, when executed, cause a computer to monitor actual rates of the branches comprise instructions that, when executed, cause a computer to average activity in the branches during a window in time; and

the instructions that, when executed, cause a computer to monitor an actual activity rate of the simulated cell body comprise instructions that, when executed, cause a computer to average activity in the simulated cell body during a window in time.

25. The computer program of claim 23 wherein:

the cell body activation threshold level comprises a number of activated branches required to activate the simulated cell body; and

the simulated cell body is configured to activate in response to a number of the simulated branched neural element's branches that are activated during a specified window of time exceeding the number of activated branches required to activate the simulated cell body.

26. The computer program of claim 23 wherein the instructions that, when executed, cause a computer program to regulate a branch's branch activation threshold level comprise instructions that, when executed, cause a computer to:

increase the branch's branch activation threshold level when the branch's actual activity rate is greater than the branch's desired activity rate; and

decrease the branch's branch activation threshold level when the branch's actual activity rate is less than the branch's desired activity rate.

27. The computer program of claim 23 wherein the instructions that, when executed, cause a computer to regulate the cell body activation threshold level comprise instructions that, when executed, cause a computer to:

increase the simulated cell body's activation threshold level when the simulated cell body's actual activity rate is greater than the desired activity rate for the simulated cell body; and

decrease the simulated cell body's activation threshold level when the simulated cell body's actual activity rate is less than the desired activity rate for the simulated cell body.

28. The computer program of claim 23 wherein:

the cell body activation threshold level comprises a number of activated branches required to activate the simulated cell body;

the simulated cell body is configured to activate in response to a number of the simulated branched neural element's branches that are activated during a specified window of time exceeding the number of activated branches required to activate the simulated cell body; and

the instructions that, when executed, cause a computer to regulate the cell body activation threshold level comprise instructions that, when executed, cause a computer to:

increase the number of activated branches required to activate the simulated cell body when the simulated cell body's actual activity rate is greater than the desired activity rate for the simulated cell body, and

decrease the number of activated branches required to activate the simulated cell body when the simulated cell body's actual activity rate is less than the desired activity rate for the simulated cell body.