Method and apparatus for neural temporal coding, learning and recognition

1. A method of neural temporal coding, comprising:

utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

selecting a reusable synapse associated with the neuron circuit based on a weight of the reusable synapse being reduced by the unsupervised learning rule to a value below a threshold; and

modifying at least one of a weight, a delay, or an afferent of the reusable synapse.

2. The method of claim 1 , further comprising:

controlling at least one of the weights, a number of synapses associated with the neuron circuit, time constants of the synapses, a firing threshold of the neuron circuit, a degree of inhibition of the neuron circuit, or a noise associated with the neuron circuit to induce initial firing rates of the neuron circuit.

3. The method of claim 1 , further comprising:

applying, by the unsupervised learning rule, simulated annealing for the adjustment of weights.

4. A method of neural temporal coding, comprising:

utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

selecting a reusable synapse associated with the neuron circuit, wherein selecting the reusable synapse is conditioned upon a time having elapsed since the synapse was reused; and

modifying at least one of a weight, a delay, or an afferent of the reusable synapse.

5. An apparatus for neural temporal coding, comprising a processing unit configured to:

utilize a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

apply a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjust, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

select a reusable synapse associated with the neuron circuit based on a weight of the reusable synapse being reduced by the unsupervised learning rule to a value below a threshold; and

modify at least one of a weight, a delay, or an afferent of the reusable synapse.

6. The apparatus of claim 5 , wherein the processing unit is further configured to:

control at least one of the weights, a number of synapses associated with the neuron circuit, time constants of the synapses, a firing threshold of the neuron circuit, a degree of inhibition of the neuron circuit, or a noise associated with the neuron circuit to induce initial firing rates of the neuron circuit.

7. The apparatus of claim 5 , wherein the processing system unit is further configured to:

apply, by the unsupervised learning rule, simulated annealing for the adjustment of weights.

8. A apparatus for neural temporal coding, comprising a processing unit configured to:

utilize a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

apply a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjust, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

select a reusable synapse associated with the neuron circuit, wherein the processing unit is configured to select the reusable synapse conditioned upon a time having elapsed since the synapse was reused; and

modify at least one of a weight, a delay, or an afferent of the reusable synapse.

9. An apparatus for neural temporal coding, comprising:

means for utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

means for applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

means for adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

means for selecting a reusable synapse associated with the neuron circuit based on a weight of the reusable synapse being reduced by the unsupervised learning rule to a value below a threshold; and

means for modifying at least one of a weight, a delay, or an afferent of the reusable synapse.

10. The apparatus of claim 9 , further comprising:

means for controlling at least one of the weights, a number of synapses associated with the neuron circuit, time constants of the synapses, a firing threshold of the neuron circuit, a degree of inhibition of the neuron circuit, or a noise associated with the neuron circuit to induce initial firing rates of the neuron circuit.

11. The apparatus of claim 9 , further comprising:

means for applying, by the unsupervised learning rule, simulated annealing for the adjustment of weights.

12. An apparatus for neural temporal coding, comprising:

means for utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

means for applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

means for adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

means for selecting a reusable synapse associated with the neuron circuit, wherein selecting the reusable synapse is conditioned upon a time having elapsed since the synapse was reused; and

means for modifying at least one of a weight, a delay, or an afferent of the reusable synapse.

13. A computer program product for neural temporal coding, comprising a non-transitory computer-readable medium comprising code for:

utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

selecting a reusable synapse associated with the neuron circuit based on a weight of the reusable synapse being reduced by the unsupervised learning rule to a value below a threshold; and

modifying at least one of a weight, a delay, or an afferent of the reusable synapse.

14. The computer program product of claim 13 , wherein the computer-readable medium further comprises code for:

controlling at least one of the weights, a number of synapses associated with the neuron circuit, time constants of the synapses, a firing threshold of the neuron circuit, a degree of inhibition of the neuron circuit, or a noise associated with the neuron circuit to induce initial firing rates of the neuron circuit.

15. The computer program product of claim 13 , wherein the computer-readable medium further comprises code for:

applying, by the unsupervised learning rule, simulated annealing for the adjustment of weights.

16. A computer program product for neural temporal coding, comprising a non-transitory computer-readable medium comprising code for:

utilizing a relative delay line abstraction which delays one or more synaptic inputs into a neuron circuit by time delays, wherein each synaptic input of the neuron circuit has a time delay measured relative to other time delays of other synaptic inputs of the neuron circuit;

applying a dynamic spiking model to determine spiking behavior of the neuron circuit based on weighted and delayed synaptic inputs of the neuron circuit;

adjusting, according to an unsupervised learning rule, weights associated with the synaptic inputs depending on timing relation of the neuron circuit spiking and the delayed synaptic inputs;

selecting a reusable synapse associated with the neuron circuit, wherein selecting the reusable synapse is conditioned upon a time having elapsed since the synapse was reused; and

modifying at least one of a weight, a delay, or an afferent of the reusable synapse.