Evolv Technology Solutions, Inc.

Myles Brundage

Risto Miikkulainen

The technology disclosed relates to neural network-based systems and methods of preparing a data object creation and recommendation database. Roughly described, it relates to, for each of a plurality of preliminary data object images, providing a representation of the image in conjunction with a respective conformity parameter indicating level of conformity of the image with a predefined goal, training a neural network system with the preliminary data object image representations in conjunction with their respective conformity parameters, to evaluate future data object image representations for conformity with the predefined goal, selecting a subset of secondary data object image representations, from a provided plurality of secondary data object image representations, in dependence upon the trained neural network system, and storing the image representations from the selected subset of secondary data object image representations in a tangible machine readable memory for use in a data object creation and recommendation system.

Data object creation and recommendation using machine learning based offline evolution

The technology disclosed relates to providing expanded data object correlations for user-generated web customizations. Roughly described, it relates to creating an initial plurality of candidate data object image representations by procreation, storing a candidate population having the plurality of candidate data object image representations, scoring each of the candidate data object image representations for conformity with a predefined goal using a trained neural network system and/or a rule-based percept filter, discarding candidate data object image representations from the candidate population, adding new data object image representations to the candidate population by procreation, iterating the scoring, the discarding, and the adding until the discarding yields a candidate pool of candidate data object image representations not yet discarded but which satisfy a convergence condition, and providing as recommendation for display to a user at least one of the data object images represented in the candidate pool.

Growing community of inventors

San Francisco, CA, United States of America

Myles Brundage