Feature Management Encoder Segmentation for Event Trends
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solutions for generating features for users and data in recommendation systems heavily depend on the accuracy of representing user and data attributes, and they struggle to effectively capture trends of events across multiple time windows.
Innovation Solution
A method for feature management that involves obtaining first and second events associated with objects, determining features for these objects using encoders, and updating the encoders based on these features and events, thereby improving the representation of attributes and event trends.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional feature generation methods are used, then the system is simple to implement, but the accuracy of feature representation and event trend capture is insufficient
Solution Approach 1:
The patent divides the feature generation process into multiple independent encoders (first encoder and second encoder) that operate on different time windows. Each encoder processes events independently within its designated time window, allowing the system to capture short-term and long-term event trends separately while maintaining overall system accuracy.
Solution Approach 2:
The patent introduces a time window dimension to the feature generation process by creating multiple encoders that operate on different temporal scales. This dimensional expansion allows the system to simultaneously capture events from both short-term and long-term perspectives, significantly improving feature representation accuracy without requiring a single complex encoder.
2Adaptability or versatility
If a single encoder is used for feature generation, then the system complexity is low, but the ability to capture event trends across multiple time windows is limited
Solution Approach 1:
The patent segments the event processing function across multiple encoders, with each encoder responsible for a specific time window. This segmentation enables the system to capture diverse event trends (both short-term and long-term) simultaneously, improving adaptability to different event patterns while keeping each individual encoder relatively simple.
Solution Approach 2:
The patent creates a universal feature generation framework where multiple encoders work together to handle different types of events across various time windows. Each encoder is designed with a common structure that can process different event types, making the system versatile and adaptable to multiple scenarios without requiring entirely separate processing pipelines.
Data Source
AI summary
There are proposed methods, devices, and computer program products for feature management. In the method, a first event associated with a first and a second object, and a second event associated with the first and second events are obtained, and a type of the first event is different from a type of the second event. A first feature of the first object is determined based on a first encoder, and a second feature of the second object is determined based on a second encoder. The first encoder is updated based on the first and second features and the first and second events. With these implementations, multiple events are used in determining the encoder for extracting the feature, and thus the encoder may have better performance in accuracy and increase performance of downstream tasks.