Class level artificial neural network

1. A process comprising:

identifying a plurality of classes in a first dataset;

generating independent artificial neural networks by creating an independent artificial neural network for each class in the plurality of classes in the first dataset such that an architecture for each independent artificial neural network optimizes performance for a single class for which the each independent artificial neural network is trained, wherein the architecture for each independent artificial neural network is distinct and is established by selecting a combination of layers;

providing all classes in the first dataset into each independent artificial neural network;

separately training each independent artificial neural network to respond to the single class in the first dataset for which the architecture of the each independent artificial neural network was optimized and to reject all other classes in the first dataset;

providing output from each independent artificial neural network to a combining classifier;

training the combining classifier to identify all classes from the first dataset based on the output of all the independent artificial neural networks;

providing a second dataset to each independent artificial neural network; and

identifying the single class in the second dataset using each independent artificial neural network such that a particular independent artificial neural network is signaled as having identified the second dataset when a value calculated by the particular independent artificial neural network for the second dataset exceeds a threshold.

2. The process of claim 1 , comprising bypassing the combining classifier when only one of the independent artificial neural networks identifies the single class in the second dataset.

3. The process of claim 1 , comprising adding a new independent artificial neural network when the combining classifier is not able to identify the single class.

4. The process of claim 1 , wherein the independent artificial neural networks are trained in parallel.

5. The process of claim 1 , comprising designing an architecture for a particular independent artificial neural network to optimize performance of the particular independent artificial neural network for the single class that the particular independent artificial neural network is trained to respond.

6. A non-transitory computer readable medium comprising instructions that when executed by a processor execute a process comprising:

identifying a plurality of classes in a first dataset;

generating independent artificial neural networks by creating an independent artificial neural network for each class in the plurality of classes in the first dataset such that an architecture for each independent artificial neural network optimizes performance for a single class for which the each independent artificial neural network is trained, wherein the architecture for each independent artificial neural network is distinct and is established by selecting a combination of layers;

providing all classes in the first dataset into each independent artificial neural network;

separately training each independent artificial neural network to respond to the single class in the first dataset for which the architecture of the each independent artificial neural network was optimized and to reject all other classes in the first dataset;

providing output from each independent artificial neural network to a combining classifier;

training the combining classifier to identify all classes from the first dataset based on the output of all the independent artificial neural networks;

providing a second dataset to each independent artificial neural network; and

identifying the single class in the second dataset using each independent artificial neural network such that a particular independent artificial neural network is signaled as having identified the second dataset when a value calculated by the particular independent artificial neural network for the second dataset exceeds a threshold.

7. The non-transitory computer readable medium of claim 6 , comprising instructions for bypassing the combining classifier when only one of the independent artificial neural networks identifies the single class in the second dataset.

8. The non-transitory computer readable medium of claim 6 , comprising instructions for adding a new independent artificial neural network when the combining classifier is not able to identify the single class.

9. The non-transitory computer readable medium of claim 6 , wherein the independent artificial neural networks are trained in parallel.

10. The non-transitory computer readable medium of claim 6 , comprising instructions for designing an architecture for a particular independent artificial neural network to optimize a performance of the particular independent artificial neural network for the single class that the particular independent artificial neural network is trained to respond.

11. A system comprising:

a computer processor; and a memory coupled to the computer processor;

wherein the computer processor and the memory are operable for:

identifying a plurality of classes in a first dataset;

generating independent artificial neural networks by creating an independent artificial neural network for each class in the plurality of classes in the first dataset such that an architecture for each independent artificial neural network optimizes performance for a single class for which the each independent artificial neural network is trained, wherein the architecture for each independent artificial neural network is distinct and is established by selecting a combination of layers;

providing all classes in the first dataset into each independent artificial neural network;

separately training each independent artificial neural network to respond to the single class in the first dataset for which the architecture of the each independent artificial neural network was optimized and to reject all other classes in the first dataset;

providing output from each independent artificial neural network to a combining classifier;

training the combining classifier to identify all classes from the first dataset in a second dataset based on the output of all the independent artificial neural networks;

providing a second dataset to each independent artificial neural network; and

identifying the single class in the second dataset using each independent artificial neural network such that a particular independent artificial neural network is signaled as having identified the second dataset when a value calculated by the particular independent artificial neural network for the second dataset exceeds a threshold.

12. The system of claim 11 , wherein the computer processor is operable for bypassing the combining classifier when only one of the independent artificial neural networks identifies the single class in the second dataset.

13. The system of claim 11 , wherein the computer processor is operable for adding a new independent artificial neural network when the combining classifier is not able to identify the single class.

14. The system of claim 11 , wherein the independent artificial neural networks are trained in parallel.

15. The system of claim 11 , wherein the processor is operable for designing an architecture for a particular independent artificial neural network to optimize a performance of the particular independent artificial neural network for the single class that the particular independent artificial neural network is trained to respond.