User classification from data via deep segmentation for semi-supervised learning

A method of user classification, comprising:

• identifying user information for each of a first set of users;
• selecting a second set of users, wherein the second set of users is a proper subset of the first set of users;
• collecting additional information for each of the second set of users;
• segmenting the second set of users into multiple categories based on the additional information;
• training a first neural network and a second neural network based on the segmentation of the second set of users, wherein the first neural network maps the user information to an embedding space and the second neural network maps the embedding space to a space of probability vectors corresponding to the multiple categories;
• mapping the user information for the first set of users to the embedding space using the first neural network;
• segmenting the first set of users into the multiple categories using a low-density separation algorithm;
• updating the first neural network and the second neural network based on the segmentation of the first set of users;
• and
• reclassifying the first set of users based on the updated first neural network and the updated second neural network.

A method of object classification, comprising:

• identifying information for a first set of objects;
• identifying additional information for a second set of objects, wherein the second set of objects is a subset of the first set of objects;
• classifying the second set of objects into multiple categories based on the additional information;
• training a first neural network and a second neural network based on the segmentation of the second set of objects, wherein the first neural network maps the object information to an embedding space and the second neural network maps the embedding space to a space corresponding to the multiple categories;
• mapping the first set of objects to the embedding space using the first neural network;
• classifying the first set of objects into the multiple categories using a low-density separation algorithm based on the mapping;
• retraining the first neural network and the second neural network based on the classification of the first set of objects;
• remapping the first set of objects to the embedding space using the updated first neural network;
• reclassifying the first set of objects into the multiple categories using the low-density separation algorithm based on the remapping;
• and
• retraining the first neural network and the second neural network based on the reclassification of the first set of objects.