Method and apparatus for large scale machine learning

1. A method for analyzing patterns in a volume of data and taking an action based on the analysis, comprising:

receiving the volume of data;

training in parallel a first plurality of Artificial Neural Networks (ANNs) of different sizes, each of the first plurality of ANNs a self-organizing map, using the data, to create a plurality of new representative data points that are representative of different classes of patterns in the data and are represented by nodes in the first plurality of ANNs;

ranking a plurality of features using the plurality of new representative data points that are representative of different classes of patterns in the data, for each of the different classes of patterns in the data;

selecting, using the ranked plurality of features, a plurality of features that are predictive of different classes of patterns in the data;

training in parallel a second plurality of ANNs of different sizes, and in different feature spaces comprising different combinations of the ranked and selected plurality of features, each of the second plurality of ANNs a self-organizing map, using the data;

extracting only active nodes that are representative of a class of patterns in the data from the second plurality of ANNs and adding class labels to each extracted active node;

determining a combination of features that are predictive of different classes of patterns in the data;

retaining the extracted class-labeled active nodes corresponding to the combination of features that are predictive of different classes of patterns in the data;

classifying patterns in the data based on the retained class-labeled active nodes; and

taking an action based on the classified patterns in the data.

2. The method of claim 1 , wherein receiving the volume of data comprises receiving streaming data (“a data stream”).

3. The method of claim 1 , further comprising discarding the first plurality of ANNs prior to training in parallel the second plurality of ANNs.

4. The method of claim 1 , wherein selecting, using the ranked plurality of features, the plurality of features that are predictive of different classes of patterns in the data, using the representative data points, reduces dimensionality of the data.

5. The method of claim 1 , wherein selecting, the ranked plurality of features, the plurality of features that are predictive of different classes of patterns in the data, using the representative data points, produces a separation between patterns in different classes and also makes the patterns within each class more compact.

6. A method for classifying data having a plurality of features belonging to a plurality of classes of patterns in streaming data, comprising:

receiving the streaming data;

training in parallel a first plurality of Artificial Neural Networks (ANNs), of different grid sizes, and each of which is a self-organizing map, by processing the streaming data, wherein the first plurality of ANNs form clusters, wherein active nodes in the first plurality of ANNs are the centers of the clusters and serve as representative examples of different classes in the streaming data;

assigning the active nodes in the first plurality of ANNs to one of the plurality of classes;

ranking a plurality of features for each of the plurality of classes;

grouping one or more of the plurality of features into each of a plurality of separate categories based on the ranking;

training in parallel a second plurality of ANNs, of different grid sizes and in different feature spaces comprising different combinations of the grouped and ranked plurality of features, and each of which is a self-organizing map, by processing the streaming data;

assigning active nodes in the second plurality of ANNs to one of the plurality of classes; and

creating a group of hypersphere-based classifiers from a subset of the active nodes in the second plurality of ANNs.

7. The method of claim 6 , wherein assigning the active nodes in the first plurality of ANNs to one of the plurality of classes is based on a number of times each of the active nodes in the first plurality of ANNs is activated by data points belonging to each of the plurality of classes.

8. The method of claim 6 , wherein assigning the active nodes in the second plurality of ANNs to one of the plurality of classes is based on a number of times each of the active nodes in the second plurality of ANNs is activated by data points belonging to each of the plurality of classes.

9. The method of claim 6 , wherein ranking the plurality of features for each of the plurality of classes is based on separability indices for each feature.

10. The method of claim 9 , further comprising computing the separability indices for each feature in the plurality of features and for each class in the plurality of classes prior to ranking the plurality of features for each of the plurality of classes based on the separability indices for each feature.

11. The method of claim 6 , wherein:

receiving the streaming data comprises receiving streaming data that is unbalanced; and

the active nodes in the first plurality of ANNs that are the centers of the clusters that serve as representative examples of different classes in the streaming data serve as representative examples of majority classes and minority classes in the streaming data; and

the method further comprising:

training in parallel a third plurality of ANNs, of different grid sizes and for a plurality of different subsets of the plurality of features, by processing the streaming data,

wherein active nodes in the third plurality of ANNs serve as representative examples of majority classes and minority classes in the streaming data when a minimum threshold is exceeded in the of active nodes in the third plurality of ANNs.

12. The method of claim 11 , wherein the representative examples of majority classes and minority classes in the streaming data are created for each class of patterns in the received streaming data and are represented by active nodes in the first plurality of ANNs.

13. A non-transitory computer-readable medium containing computer-executable instructions that, when executed by a processor, cause the processor to analyze patterns in a volume of data and take an action based on the analysis, according to a method comprising:

receiving the volume of data;

training in parallel a first plurality of Artificial Neural Networks (ANNs) of different sizes, each of the first plurality of ANNs a self-organizing map, using the data, to create a plurality of new representative data points that are representative of different classes of patterns in the data and are represented by nodes in the first plurality of ANNs;

ranking a plurality of features using the plurality of new representative data points that are representative of different classes of patterns in the data, for each of the different classes of patterns in the data;

selecting, using the ranked plurality of features, a plurality of features that are predictive of different classes of patterns in the data;

training in parallel a second plurality of ANNs of different sizes, and in different feature spaces comprising different combinations of the ranked and selected plurality of features, each of the second plurality of ANNs a self-organizing map, using the data;

extracting only active nodes that are representative of a class of patterns in the data from the second plurality of ANNs and adding class labels to each extracted active node;

determining a combination of features that are predictive of different classes of patterns in the data;

retaining the extracted class-labeled active nodes corresponding to the combination of features that are predictive of different classes of patterns in the data;

classifying patterns in the data based on the retained class-labeled active nodes; and

taking an action based on the classified patterns in the data.

14. The non-transitory computer-readable medium of claim 13 , further comprising discarding the first plurality of ANNs prior to training in parallel the second plurality of ANNs.

15. The non-transitory computer-readable medium of claim 13 wherein selecting, using the ranked plurality of features, the plurality of features that are predictive of different classes of patterns in the data, using the representative data points, reduces dimensionality of the data.

16. The non-transitory computer-readable medium of claim 13 wherein selecting, using the ranked plurality of features, the plurality of features that are predictive of different classes of patterns in the data, using the representative data points, produces a separation between patterns in different classes and also makes the patterns within each class more compact.