Devices and methods for increasing the speed or power efficiency of a computer when performing machine learning using spiking neural networks
A method for increasing a speed and efficiency of a computer when performing machine learning using spiking neural networks. The method includes computer-implemented operations; that is, operations that are solely executed on a computer. The method includes receiving, in a spiking neural network, a plurality of input values upon which a machine learning algorithm is based. The method also includes correlating, for each input value, a corresponding response speed of a corresponding neuron to a corresponding equivalence relationship between the input value to a corresponding latency of the corresponding neuron. Neurons that trigger faster than other neurons represent close relationships between input values and neuron latencies. Latencies of the neurons represent data points used in performing the machine learning. A plurality of equivalence relationships are formed as a result of correlating. The method includes performing the machine learning using the plurality of equivalence relationships.
1. A method for increasing a speed and efficiency of a computer when performing machine learning using spiking neural networks, the method comprising computer-implemented operations of:
receiving, in a spiking neural network, a plurality of input values upon which a machine learning algorithm is based;
correlating, for each input value, a corresponding response speed of a corresponding neuron to a corresponding equivalence relationship between the input value to a corresponding latency of the corresponding neuron, wherein neurons that trigger faster than other neurons represent close relationships between input values and neuron latencies, wherein latencies of the neurons represent data points used in performing the machine learning, and wherein a plurality of equivalence relationships are formed as a result of correlating; and
performing the machine learning using the plurality of equivalence relationships.
2. The method of claim 1 wherein the machine learning comprises a k-nearest neighbor algorithm.
3. The method of claim 2 further comprising:
performing a spiking similarity function and identifying the first k spike responses corresponding to the k most similar data points to an uncategorized query point.
4. The method of claim 3 further comprising:
determining a k nearest neighbors from a k-winner layer; and
priming a max layer to determine a class of a majority of the k winning nearest neighbors.
5. The method of claim 4 further comprising:
feeding, from the max layer, each class neuron unit input until a majority class fires, thereby yielding a classification.
6. The method of claim 1 wherein the machine learning comprises an adaptive resonance theory learning classifier.
7. The method of claim 6 further comprising:
comparing inputs against stored templates to determine if a sufficiency similar representation defined by a vigilance parameter exists or if a new category is to be learned.
8. The method of claim 1 wherein the machine learning comprises a support vector machine.
9. The method of claim 8 further comprising:
determining, for each data point, a distance of a given data point to all data points of an opposing class, wherein a set of distances are created;
aggregating a set of closest points from the set of distances; and
reading out support vectors from the set of closest points using a k-winner accumulator.
10. A non-transitory computer readable storage medium storing program code which, when executed by a processor, performs a computer-implemented method for increasing a speed and efficiency of a computer when performing machine learning using spiking neural networks, the program code comprising:
computer usable program code for receiving, in a spiking neural network, a plurality of input values upon which a machine learning algorithm is based;
computer usable program code for correlating, for each input value, a corresponding response speed of a corresponding neuron to a corresponding equivalence relationship between the input value to a corresponding latency of the corresponding neuron, wherein neurons that trigger faster than other neurons represent close relationships between input values and neuron latencies, wherein latencies of the neurons represent data points used in performing the machine learning, and wherein a plurality of equivalence relationships are formed as a result of correlating; and
computer usable program code for performing the machine learning using the plurality of equivalence relationships.
11. The non-transitory computer readable storage medium of claim 10 wherein the machine learning comprises a k-nearest neighbor algorithm.
12. The non-transitory computer readable storage medium of claim 11 wherein the program code further comprises:
computer usable program code for performing a spiking similarity function and identifying the first k spike responses corresponding to the k most similar data points to an uncategorized query point;
computer usable program code for determining a k nearest neighbors from a k-winner layer;
computer usable program code for priming a max layer to determine a class of a majority of the k winning nearest neighbors; and
computer usable program code for feeding, from the max layer, each class neuron unit input until a majority class fires, thereby yielding a classification.
13. The non-transitory computer readable storage medium of claim 10 wherein the machine learning comprises an adaptive resonance theory learning classifier, and wherein the program code further comprises:
computer usable program code for comparing inputs against stored templates to determine if a sufficiency similar representation defined by a vigilance parameter exists or if a new category is to be learned.
14. The non-transitory computer readable storage medium of claim 10 wherein the machine learning comprises a support vector machine, and wherein the program code further comprises:
program code for determining, for each data point, a distance of a given data point to all data points of an opposing class, wherein a set of distances are created;
program code for aggregating a set of closest points from the set of distances; and
program code for reading out support vectors from the set of closest points using a k-winner accumulator.
15. A computer comprising:
a processor; and
a non-transitory computer readable storage medium in communication with the processor and storing program code which, when executed by the processor, performs a computer-implemented method for increasing a speed and efficiency of a computer when performing machine learning using spiking neural networks, the program code comprising:
computer usable program code for receiving, in a spiking neural network, a plurality of input values upon which a machine learning algorithm is based;
computer usable program code for correlating, for each input value, a corresponding response speed of a corresponding neuron to a corresponding equivalence relationship between the input value to a corresponding latency of the corresponding neuron, wherein neurons that trigger faster than other neurons represent close relationships between input values and neuron latencies, wherein latencies of the neurons represent data points used in performing the machine learning, and wherein a plurality of equivalence relationships are formed as a result of correlating; and
computer usable program code for performing the machine learning using the plurality of equivalence relationships.
16. The computer of claim 15 wherein the machine learning comprises a k-nearest neighbor algorithm and wherein the program code further comprises:
computer usable program code for performing a spiking similarity function and identifying the first k spike responses corresponding to the k most similar data points to an uncategorized query point;
computer usable program code for determining a k nearest neighbors from a k-winner layer;
computer usable program code for priming a max layer to determine a class of a majority of the k winning nearest neighbors; and
computer usable program code for feeding, from the max layer, each class neuron unit input until a majority class fires, thereby yielding a classification.
17. The computer of claim 15 wherein the machine learning comprises an adaptive resonance theory learning classifier.
18. The computer of claim 17 wherein the program code further comprises:
computer usable program code for comparing inputs against stored templates to determine if a sufficiency similar representation defined by a vigilance parameter exists or if a new category is to be learned.
19. The computer of claim 15 wherein the machine learning comprises a support vector machine.
20. The computer of claim 19 wherein the program code further comprises:
program code for determining, for each data point, a distance of a given data point to all data points of an opposing class, wherein a set of distances are created;
program code for aggregating a set of closest points from the set of distances; and
program code for reading out support vectors from the set of closest points using a k-winner accumulator.