Selective Breeding for Divergent Edge Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In edge computing environments, neural networks deployed on edge servers may drift and diverge from one another, leading to reduced performance when encountering inputs not seen at specific edge nodes, while existing methods fail to efficiently update and breed these neural networks to adapt to local data.
Innovation Solution
Deploying multiple copies of a centralized neural network to edge servers, where each copy is independently trained and updated, with periodic sharing and breeding of neural networks across servers to adapt to local data and improve performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If neural networks are independently trained at each edge server, then adaptability to local data is improved, but divergence between networks increases
Solution Approach 1:
The system segments the neural network training process into independent local training at each edge server while maintaining periodic synchronization. Each edge server trains its neural network independently on local data, but periodically exchanges models with other servers to prevent divergence, creating a distributed training architecture that balances autonomy with coherence.
Solution Approach 2:
The system implements feedback mechanisms where edge servers periodically receive updated neural network models from other servers and integrate these feedback models through breeding operations. This continuous feedback loop allows networks to adapt to local data while maintaining similarity to other networks in the system.
2Reliability
If neural networks are updated frequently across edge servers, then performance is improved, but network traffic increases
Solution Approach 1:
The system employs periodic action by updating neural networks at predetermined time intervals rather than continuously. Edge servers wait for scheduled update cycles to exchange models, reducing the frequency of communications while still maintaining performance improvements through regular synchronization.
Solution Approach 2:
The system uses copying by distributing identical initial neural network models to all edge servers and exchanging copied models periodically. This allows performance improvement through model sharing without requiring constant data transmission, as only model copies are exchanged at intervals rather than continuous data streams.
3Adaptability or versatility
If edge servers share trained neural networks, then adaptability is improved, but device complexity increases
Solution Approach 1:
The system merges neural network models from multiple edge servers through breeding operations that combine characteristics of parent networks. This merging process creates improved offspring models that inherit adaptive capabilities from multiple sources while consolidating complexity into unified models rather than requiring separate management of multiple networks.
Solution Approach 2:
The system applies parameter changes by modifying neural network parameters through breeding operations that select and recombine parameters from parent networks. This parameter-level manipulation allows adaptability improvement through model exchange without requiring complex system architecture, as the breeding process operates directly on network parameters rather than requiring complex coordination mechanisms.
Data Source
AI summary
Selective breeding for divergent neural networks in an edge computing environment includes deploying a plurality of copies of a centralized neural network respectively to a corresponding plurality of edge servers, wherein each of the copies of the centralized neural network is independently operated and trained at one of the edge servers based on inputs received at that edge server and becomes an independently trained neural network. Each of the edge servers at periodic intervals sends a copy of the independently trained neural network at that edge server to other ones of the edge servers. At each of one or more of the edge servers, the independently trained neural network at that edge server is updated, including performing neural network breeding based on the independently trained neural network at that edge server and one or more copies of the independently trained neural networks sent to the edge server.