18626091. USING SHARED AND NON-SHARED PARAMETERS IN AN ATTENTION MODULE-BASED RECOGNITION MODEL simplified abstract (TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED)
Contents
USING SHARED AND NON-SHARED PARAMETERS IN AN ATTENTION MODULE-BASED RECOGNITION MODEL
Organization Name
TENCENT TECHNOLOGY (SHENZHEN) COMPANY LIMITED
Inventor(s)
Shidong Shang of Shenzhen (CN)
USING SHARED AND NON-SHARED PARAMETERS IN AN ATTENTION MODULE-BASED RECOGNITION MODEL - A simplified explanation of the abstract
This abstract first appeared for US patent application 18626091 titled 'USING SHARED AND NON-SHARED PARAMETERS IN AN ATTENTION MODULE-BASED RECOGNITION MODEL
The recognition method described in the patent application involves inputting a media resource feature into a recognition model with N layers of attention modules to obtain a representation vector.
- An ilayer of attention module determines weight parameters and input representation vectors based on shared and non-shared parameters.
- The ilayer of attention module outputs a representation vector used by subsequent layers of attention modules.
- Shared parameters are utilized by at least two layers of attention modules among the N layers.
Potential Applications: - Image recognition systems - Video analysis software - Speech recognition technology
Problems Solved: - Enhances the accuracy of recognition models - Improves the efficiency of processing media resources
Benefits: - Higher precision in recognizing media features - Faster processing of large datasets - Enhanced performance of recognition models
Commercial Applications: Title: Advanced Media Recognition Technology for Enhanced Data Analysis This technology can be used in various industries such as: - Security and surveillance - Marketing and advertising - Healthcare diagnostics
Questions about Media Recognition Technology: 1. How does this technology improve the accuracy of recognition models?
This technology utilizes multiple layers of attention modules to extract key features from media resources, leading to more precise recognition results.
2. What are the potential implications of this technology in the field of artificial intelligence?
This technology can significantly enhance the capabilities of AI systems in various applications such as image and speech recognition.
Original Abstract Submitted
A recognition method includes inputting a media resource feature into a recognition model having N layers of attention modules, and processing the media resource feature by using the N layers of attention modules to obtain a representation vector. An ilayer of attention module is configured to determine an ilayer weight parameter and an iinput representation vector based on (i) a group of shared parameters and (ii) an igroup of non-shared parameters, and determine an ilayer of representation vector outputted by the ilayer of attention module. When i is less than N, the ilayer of representation vector determines an (i+1)group of non-shared parameters used by an (i+1)layer of attention module and at least two layers of attention modules among the N layers of attention modules share the group of shared parameters.
