Adobe, Inc.

Alexandru Ionut Hodorogea

Ryan A Rossi

Anup Bandigadi Rao

Ritwik Sinha

Tung Mai

Matvey Kapilevich

Yasin Abbasi Yadkori

Virgil-Artimon Palanciuc

This disclosure relates to methods, non-transitory computer readable media, and systems that, upon request for a trait-intersection count of users (or other digital entities) corresponding to traits for a target time period, use a machine-learning model to analyze a semantic-trait embedding of the traits and to generate an estimated trait-intersection count of such entities sharing the traits for the target time period. By applying a machine-learning model trained to estimate trait-intersection counts, the disclosed methods, non-transitory computer readable media, and systems can analyze both a semantic-trait embedding of traits and an initial trait-intersection count of trait-sharing entities for an initial time period to estimate the trait-intersection count for the target time period. The disclosed machine-learning model can thus analyze both the semantic-trait embedding and the initial trait-intersection count to efficiently and accurately estimate a trait-intersection count corresponding to a requested time period.

Generating estimated trait-intersection counts utilizing semantic-trait embeddings and machine learning

The present disclosure relates to training a recommendation model to generate trait recommendations using one permutation hashing and populated-value-slot-based densification. In particular, the disclosed systems can train the recommendation model by computing sketch vectors corresponding to traits using one permutation hashing. The disclosed systems can then fill in unpopulated value slots of the sketch vectors using populated-value-slot-based densification. The disclosed systems can combine the resulting densified sketches to generate the trained recommendation model. For example, in some embodiments, the disclosed systems can combine the sketches by generating a plurality of locality sensitive hashing tables based on the sketches. In some embodiments, the disclosed systems generate a count sketch matrix based on the sketches and generate trait embeddings based on the count sketch matrix using spectral embedding. Based on the trait embeddings, the disclosed systems can utilize the recommendation model to flexibly and accurately determine the similarity between traits.

Utilizing one hash permutation and populated-value-slot-based densification for generating audience segment trait recommendations

Growing community of inventors

Bucharest, Romania

Alexandru Ionut Hodorogea