18148252. INTEGRATED ARCHITECTURE SEARCHING SYSTEM AND METHOD simplified abstract (TSINGHUA UNIVERSITY)
Contents
INTEGRATED ARCHITECTURE SEARCHING SYSTEM AND METHOD
Organization Name
Inventor(s)
INTEGRATED ARCHITECTURE SEARCHING SYSTEM AND METHOD - A simplified explanation of the abstract
This abstract first appeared for US patent application 18148252 titled 'INTEGRATED ARCHITECTURE SEARCHING SYSTEM AND METHOD
Simplified Explanation: The patent application describes an integrated architecture searching system for a click-through rate prediction model. This system involves searching for embedding vector dimensions of features, determining feature interactions, and predicting click-through rates using deep networks.
- **Feature Embedding:** Search for embedding vector dimensions of features to match pairs of features.
- **Feature Interaction:** Obtain feature interaction results by searching for feature interaction sub-networks and combinations.
- **Click-Through Rate Prediction:** Incorporate feature interaction results into high-order implicit feature interaction search space and perform high-order implicit feature interaction on deep networks of different layers.
Key Features and Innovation:
- Search for embedding vector dimensions of features.
- Determine feature interactions through sub-networks and combinations.
- Predict click-through rates using high-order implicit feature interaction on deep networks.
Potential Applications: This technology can be applied in online advertising, recommendation systems, and personalized content delivery.
Problems Solved: This technology addresses the challenge of accurately predicting click-through rates by incorporating feature interactions in deep networks.
Benefits:
- Improved accuracy in click-through rate prediction.
- Enhanced performance of recommendation systems.
- Personalized content delivery for users.
Commercial Applications: The technology can be utilized in digital marketing, e-commerce platforms, and online content platforms to optimize user engagement and increase click-through rates.
Prior Art: Researchers can explore prior studies on feature interactions in deep learning models and click-through rate prediction to understand the existing knowledge in this field.
Frequently Updated Research: Stay updated on advancements in deep learning models for click-through rate prediction and feature interaction analysis to enhance the performance of this technology.
Questions about Click-Through Rate Prediction: 1. How does the system determine the matching embedding vector dimension for each pair of features? 2. What are the potential challenges in incorporating high-order implicit feature interaction in deep networks for click-through rate prediction?
Original Abstract Submitted
An integrated architecture searching system for a click-through rate prediction model is provided. The system includes a first search space configured to search for embedding vector dimensions of features and determine a matching embedding vector dimension for each pair of the features; a second search space configured to obtain a feature interaction result by searching for a feature interaction sub-network and a feature interaction combination; and a third search space configured to obtain a click-through rate prediction value by incorporating the feature interaction result into an high-order implicit feature interaction search space and performing high-order implicit feature interaction on deep networks of different layers.
