Adaptive Information Sharing Levels for Distributed Model Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to optimally determine the amount of shared information between devices in a device group for maximizing overall performance.
Innovation Solution
A communication method that dynamically adjusts the information sharing level between devices based on specific sharing levels, allowing devices to control the amount of shared information accurately, thereby optimizing overall performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If devices share a fixed amount of information, then the system structure is simple, but the overall performance cannot be continuously optimized
Solution Approach 1:
The patent implements dynamic information sharing levels that can be adjusted over time. Devices transition between different information sharing levels (first level, second level, etc.) based on performance feedback, allowing the system to adapt and optimize overall performance while maintaining manageable complexity through structured level transitions
Solution Approach 2:
The patent employs feedback mechanisms where devices send performance information to a third device, which then determines appropriate information sharing levels. This closed-loop feedback system enables continuous performance optimization by adjusting information sharing amounts based on actual system performance, resolving the contradiction between adaptability and complexity
2Productivity
If devices dynamically adjust information sharing amount, then overall performance is optimized, but the control mechanism becomes complex
Solution Approach 1:
The patent segments the information sharing process into distinct levels (first information sharing level, second information sharing level, etc.), each with specific shared information types. This segmentation allows devices to efficiently adjust information sharing amounts for optimized model training while managing complexity through structured, discrete level transitions rather than continuous control
Solution Approach 2:
The patent changes the parameter of information sharing amount by transitioning between predefined sharing levels. Each level corresponds to a specific amount and type of shared information, enabling efficient model training optimization while managing system complexity through parameter discretization and structured level-based control
3Use of energy by moving object
If a third device determines information sharing levels, then power consumption of individual devices is reduced, but system complexity increases
Solution Approach 1:
The patent introduces a third device as an intermediary that receives performance information from devices and determines appropriate information sharing levels. This intermediary approach reduces the computational burden and power consumption of individual devices while managing system complexity through a dedicated central controller that handles the complex decision-making logic
Data Source
AI summary
In an information sharing method, a communication apparatus obtains first information sharing level of a first device, and sends first shared information corresponding to the first information sharing level to a second device. The apparatus obtains a second information sharing level of the first device, and sends second shared information corresponding to the second information sharing level to the second device. The second information sharing level is different from the first information sharing level, the second shared information is not exactly the same as the first shared information, and the first shared information and the second shared information are for model training.