Convex optimized stochastic vector sampling based representation of ground truth

A system comprising:

• processing circuitry;
• and
• a non-transitory computer-readable medium, coupled to the processing circuitry, having stored therein instructions that when executed by the processing circuitry cause the processing circuitry to:
• receive a first set of feature vectors, wherein the first set of feature vectors are un-labeled;
• homomorphically translate the first set feature vectors using a T-Distributed Stochastic Neighbor Embedding (t-SNE) process to obtain a second set of feature vectors with reduced dimensionality;
• cluster the second set of feature vectors to obtain an initial set of clusters using centroid-based clustering;
• identify an optimal set of clusters among the initial set of clusters by performing a convex optimization process on the initial set of clusters;
• for each cluster of the optimal set of clusters, select a representative vector from the cluster for labeling;
• create a set of labeled feature vectors, based at least in part on one or more labels received from an oracle, for the representative vector of each cluster;
• select, by the processing circuitry, one or more boundary condition feature vectors from the first set of feature vectors for labeling by the oracle;
• train a machine learning model for multi-class classification based at least in part on the set of labeled feature vectors and the one or more boundary condition feature vectors as labeled by the oracle;
• and
• execute, by the processing circuitry, the machine learning model to classify a plurality of computing devices based at least in part on the first set of feature vectors.

A non-transitory machine readable medium storing instructions that when executed by processing circuitry of a computer system cause the processing circuitry to:

• receive a first set of feature vectors, wherein the first set of feature vectors are un-labeled;
• homomorphically translate the first set feature vectors using a T-Distributed Stochastic Neighbor Embedding (t-SNE) process to obtain a second set of feature vectors with reduced dimensionality;
• cluster the second set of feature vectors to obtain an initial set of clusters using centroid-based clustering;
• identify an optimal set of clusters among the initial set of clusters by performing a convex optimization process on the initial set of clusters;
• and
• for each cluster of the optimal set of clusters, select a representative vector from the cluster for labeling;
• create a set of labeled feature vectors, based at least in part on one or more labels received from an oracle, for the representative vector of each cluster;
• select, by the processing circuitry, one or more boundary condition feature vectors from the first set of feature vectors for labeling by the oracle;
• train a machine learning model for multi-class classification based at least in part on the set of labeled feature vectors and the one or more boundary condition feature vectors as labeled by the oracle;
• and
• execute, by the processing circuitry, the machine learning model to classify a plurality of computing devices based at least in part on the first set of feature vectors.