AI Model Refinement via Gradient Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The refinement of artificial intelligence (AI) models in electronic devices poses a risk of sensitive user information leakage during transmission, particularly when context information containing personal, biometric, or financial data is shared, highlighting the need for secure and efficient methods to update AI models without compromising user privacy.
Innovation Solution
An electronic apparatus and server system that detects changes in context information, determines a gradient for refining a local AI model, transmits this gradient instead of raw context data to the server, and receives refined global model information to update the local model, thereby minimizing data transmission and protecting sensitive user information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If context information containing sensitive user data is transmitted to the server for AI model refinement, then the AI model can be continuously improved based on real-time context changes, but the risk of data leakage and privacy violation increases
Solution Approach 1:
The patent extracts only the necessary gradient information from the context data for model refinement, leaving the sensitive raw context information local to the electronic device. This extraction approach allows the server to receive minimal data (gradients) that are insufficient for reconstructing original user data, thus reducing data leakage risk while maintaining model adaptability
Solution Approach 2:
Instead of transmitting context information to the server for processing, the patent inverts the approach by having the electronic device perform local model refinement using context information, then transmit only the refinement results (gradients) to the server. This inversion eliminates the need to share sensitive context data while achieving the same model improvement goal
2Productivity
If raw context information is transmitted to the server for model refinement, then the server can perform comprehensive learning, but the data transmission volume and processing burden increase
Solution Approach 1:
The patent extracts only the essential gradient information needed for model refinement, discarding the voluminous raw context data. This extraction reduces data transmission volume from potentially large context datasets to compact gradient representations, while the server receives sufficient information to perform effective model updates
Solution Approach 2:
The patent inverts the traditional approach by performing the computationally intensive processing locally at the electronic device, then transmitting only the refined results to the server. This inversion shifts the processing burden from the server to the device, reducing network traffic while maintaining refinement quality
Data Source
AI summary
Provided are an artificial intelligence (AI) system simulating a function of a human brain, such as cognition and judgment, using a machine learning algorithm, such as deep learning, and an application thereof. Also, provided is a method, performed by an electronic apparatus, of refining an artificial intelligence (AI) model, the method including: detecting information about a context of an electronic apparatus used to refine a local model stored in the electronic apparatus being changed; determining a gradient for refining the local model based on the changed information about the context; refining the local model based on the determined gradient; transmitting the gradient to a server; receiving, from the server, information about a global model refined based on the gradient; and refining the local model based on the received information.