Space efficient random forests implementation utilizing automata processors

A method of computing a Random Forest model, comprising:

• receiving a feature vector of a data stream at a processing resource, wherein the feature vector comprises a set of feature values;
• calculating a set of feature labels based at least in part on the set of feature values to generate a label vector;
• providing the label vector to another processing resource;
• receiving a plurality of classifications corresponding to each feature label of the label vector from the other processing resource, wherein the plurality of classifications are generated based at least in part on a respective range of feature values of the set of feature values;
• and
• combining the plurality of classifications to generate a final classification of the data stream.

A non-transitory computer-readable medium having computer executable code stored thereon, the code comprising instructions to:

• cause a processing resource to receive a feature vector of a data stream, wherein the feature vector comprises a set of feature values;
• cause the processing resource to calculate a set of feature labels based at least in part on the set of feature values to generate a label vector;
• cause the processing resource to provide the label vector to another processing resource;
• cause the processing resource to receive a plurality of classifications corresponding to each feature label of the label vector generated based at least in part on a respective range of feature values of the set of feature values;
• and
• cause the processing resource to combine the plurality of classifications to generate a final classification of the data stream.

An apparatus, comprising:

• a processing resource configured to:
• compute a Random Forest model to be executed on another processing resource, wherein the Random Forest model comprises a plurality of decision trees;
• convert the plurality of decision trees of the Random Forest model, wherein converting the plurality of decision trees comprises converting each of a plurality of path-to-nodes of the plurality of decision trees into a plurality of chains of decision trees based at least in part on a range of feature values;
• and
• provide the converted Random Forest model to the other processing resource to be executed by the other processing resource.

A method, comprising:

• computing, via a first processing resource, a Random Forest model to be executed on a second processing resource, wherein the Random Forest model comprises a plurality of decision trees;
• converting, via the first processing resource, the plurality of decision trees of the Random Forest model, wherein converting the plurality of decision trees comprises converting each of a plurality of path-to-nodes of the plurality of decision trees into a plurality of chains of decision trees based at least in part on a range of feature values;
• and
• providing, via the first processing resource, the converted Random Forest model to the second processing resource to be executed by the second processing resource.