Edge Neural Network Breeding for Server Deployment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In edge computing environments, deploying a neural network to a new edge server is challenging due to the difficulty in determining which existing trained edge neural networks are optimal for the new location, as edge nodes may drift and diverge from one another during individual training.
Innovation Solution
The solution involves deploying multiple copies of a centralized neural network to edge servers, where each copy is independently trained and operated based on local inputs. A new edge neural network is generated for the new edge server by performing neural network breeding using independently trained neural networks and edge devices information, along with anticipated edge devices information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple copies of centralized neural network are deployed to edge servers for independent training, then adaptability to local conditions is improved, but device complexity increases
Solution Approach 1:
The system segments the centralized neural network into multiple independent copies deployed across different edge servers. Each copy is trained independently on local data, allowing the system to adapt to diverse local conditions while maintaining manageable complexity at each node through modular deployment
Solution Approach 2:
The patent implements a nested structure where independently trained edge neural networks are combined through breeding operations to create new edge neural networks. The new networks nest characteristics from multiple parent networks, allowing adaptability to accumulate while keeping individual network complexity controlled
2Reliability
If edge neural networks are independently trained at each edge server, then performance at specific locations is improved, but difficulty in selecting optimal network for new servers increases
Solution Approach 1:
The system implements feedback through breeding operations where the performance and characteristics of independently trained networks are evaluated and used to inform the creation of new networks. This feedback loop enables systematic selection and combination of successful network traits without requiring manual evaluation of each network
Solution Approach 2:
The patent performs preliminary breeding operations using historical performance data and edge device information before deploying networks to new servers. This preliminary action pre-optimizes networks for specific locations based on anticipated device characteristics, reducing the difficulty of selection at deployment time
3Reliability
If neural networks are customized for each edge server location, then immediate performance is improved, but network congestion and latency may worsen
Solution Approach 1:
The system applies local quality by customizing neural networks for specific edge server locations based on local edge device information and training data. Each network is optimized for its local environment, ensuring immediate performance while keeping processing distributed at the edge to minimize network latency
Data Source
AI summary
Deploying a neural network to a new edge server 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 a respective one of the edge servers based on inputs received at that edge server to create independently trained neural networks. Each of the edge servers stores edge devices information including a physical location of each edge device that accessed the independently trained neural network operating at the edge server. A new edge neural network is generated for deployment to the new edge server, including performing neural network breeding based on the independently trained neural networks and the stored edge devices information, and based on anticipated edge devices information for edge devices expected to access the new edge server.