Microsoft technology licensing, llc (20240202280). LINEAR-PROGRAMMING-BASED RECOMMENDER WITH PERSONALIZED DIVERSITY CONSTRAINTS simplified abstract

LINEAR-PROGRAMMING-BASED RECOMMENDER WITH PERSONALIZED DIVERSITY CONSTRAINTS

Organization Name

microsoft technology licensing, llc

Inventor(s)

Miao Cheng of Sunnyvale CA (US)

Kinjal Basu of Stanford CA (US)

Aman Gupta of San Jose CA (US)

Sathiya K. Selvaraj of Sunnyvale CA (US)

Rahul Mazumder of Sommerville MA (US)

Haichao Wei of Santa Clara CA (US)

Haoyue Wang of Cambridge MA (US)

LINEAR-PROGRAMMING-BASED RECOMMENDER WITH PERSONALIZED DIVERSITY CONSTRAINTS - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240202280 titled 'LINEAR-PROGRAMMING-BASED RECOMMENDER WITH PERSONALIZED DIVERSITY CONSTRAINTS

The structured linear program described in the patent application is designed for use in recommender models, specifically for solving recommendation problems with diversity constraints in real-time.

• The structured linear program can handle large datasets and produce real-time results.
• It first reduces a two-sided diversity constraint to a one-sided constraint and introduces a dual variable for a constraint.
• The program defines a dual objective function and solves it using a bisection method.
• A primal solution is then recovered from the solved dual objective function, providing a set of recommended content items that satisfy the diversity constraint in real-time.

Potential Applications: - E-commerce platforms for personalized product recommendations - Streaming services for personalized content suggestions - Social media platforms for personalized friend or content recommendations

Problems Solved: - Efficiently handling diversity constraints in recommender systems - Providing real-time recommendations for large datasets

Benefits: - Improved user experience with personalized recommendations - Increased user engagement and satisfaction - Enhanced content discovery and diversity

Commercial Applications: Title: Real-Time Personalized Recommendation System This technology can be used in various commercial applications such as e-commerce platforms, streaming services, and social media platforms to enhance user experience and engagement through personalized recommendations.

Questions about the Structured Linear Program: 1. How does the structured linear program handle diversity constraints in real-time? The structured linear program reduces a two-sided diversity constraint to a one-sided constraint and introduces a dual variable to define a dual objective function, which is then solved using a bisection method to provide real-time recommendations that satisfy the diversity constraint.

2. What are the potential benefits of using the structured linear program in recommender models? The structured linear program can improve user experience by providing personalized recommendations, increase user engagement and satisfaction, and enhance content discovery and diversity.

Original Abstract Submitted

in an example embodiment, a structured linear program is provided that is usable in recommender models. this structured linear program is able to produce real-time results for a structured recommendation problem with diversity constraints, even for large data sets. the structured linear program operates by first reducing a two-sided diversity constraint to a one-sided diversity constraint, and then introducing a dual variable for a constraint, in order to define a dual objective function. the dual objective function is then solved using a bisection method. a primal solution is then recovered using the solved dual objective function. the resultant primal solution reflects a set of recommended content items that satisfy the diversity constraint, as computed in real-time.