18048428. ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING simplified abstract (Microsoft Technology Licensing, LLC)
Contents
- 1 ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Original Abstract Submitted
ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING
Organization Name
Microsoft Technology Licensing, LLC
Inventor(s)
Liyan Fang of Sunnyvale CA (US)
Andrew O. Hatch of Berkeley CA (US)
Keqing Liang of Cupertino CA (US)
Yafei Wei of Sunnyvale CA (US)
Ankan Saha of San Francisco CA (US)
ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING - A simplified explanation of the abstract
This abstract first appeared for US patent application 18048428 titled 'ENTITY SELECTION AND RANKING USING DISTRIBUTION SAMPLING
Simplified Explanation
Embodiments of the disclosed technologies include generating a reward score for an entity. A rate distribution is determined using the reward score and a number of times the entity has been selected for ranking. A sampled rate value is generated by sampling the rate distribution. A probability score is generated for a pair of the entity and a user based on the sampled rate value. A probability distribution is determined using the probability score. A sampled probability value is generated by sampling the probability distribution. A machine learning model is trained using the sampled probability value.
- Generating reward scores for entities
- Determining rate distributions based on reward scores and selection frequency
- Generating sampled rate values by sampling rate distributions
- Generating probability scores for entity-user pairs
- Determining probability distributions using probability scores
- Generating sampled probability values by sampling probability distributions
- Training machine learning models using sampled probability values
Potential Applications
This technology could be applied in recommendation systems, personalized advertising, and content ranking algorithms.
Problems Solved
This technology helps in improving user engagement, increasing click-through rates, and enhancing user experience by providing more relevant content.
Benefits
The benefits of this technology include better targeting of users, increased user satisfaction, and improved overall performance of recommendation systems.
Potential Commercial Applications
Optimizing online advertising campaigns for better ROI
Possible Prior Art
One possible prior art could be collaborative filtering algorithms used in recommendation systems.
What is the impact of this technology on user privacy?
This technology may raise concerns about user privacy as it involves collecting and analyzing user data to generate personalized recommendations.
How does this technology compare to existing recommendation systems?
This technology improves upon existing recommendation systems by incorporating machine learning models trained on sampled probability values, leading to more accurate and personalized recommendations.
Original Abstract Submitted
Embodiments of the disclosed technologies include generating a reward score for an entity. A rate distribution is determined using the reward score and a number of times the entity has been selected for ranking. A sampled rate value is generated by sampling the rate distribution. A probability score is generated for a pair of the entity and a user based on the sampled rate value. A probability distribution is determined using the probability score. A sampled probability value is generated by sampling the probability distribution. A machine learning model is trained using the sampled probability value.