Heterogeneous Edge Device Model Mixing via Leader Election
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning techniques are insufficient for handling the volume and velocity of data in heterogeneous environments, where variations in edge devices and connectivity are common, and cannot adequately process data streams from diverse devices with changing conditions.
Innovation Solution
The proposed system includes edge devices with a communication module, data collection device, machine learning module, group determination module, leader election module, and model mixing module, which analyze data, determine group membership, elect leaders, and mix local models to create a mixed model for improved data processing and model updates across heterogeneous groups of edge devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional machine learning techniques are used to process data, then data processing can be performed, but the system cannot adequately handle the volume and velocity of data in heterogeneous environments
Solution Approach 1:
The system segments the heterogeneous edge device network into multiple groups based on device characteristics and data types. Each group is further divided into clusters with elected leader devices. This hierarchical segmentation allows the system to manage large volumes of data from diverse devices by processing them in organized, manageable units rather than as a monolithic system.
Solution Approach 2:
The patent introduces a new dimension of organization by creating hierarchical groupings and clusters across the heterogeneous network. Instead of treating all devices uniformly, the system adds organizational layers (groups → clusters → leaders) that enable scalable processing of high-velocity data streams while maintaining adaptability to device diversity.
2Reliability
If a centralized approach is used to manage machine learning models, then model consistency can be maintained, but the system cannot efficiently handle real-time data processing in distributed heterogeneous environments
Solution Approach 1:
The system merges centralized model management with distributed processing by having leader devices in each cluster maintain and update local models. These local models are then shared and synchronized across the broader network, combining the reliability of centralized consistency with the speed of distributed real-time processing.
Solution Approach 2:
Leader devices act as intermediaries between individual edge devices and the broader network. They collect data from cluster members, perform local model training, and distribute updated models to other devices, enabling real-time processing while maintaining model consistency across the heterogeneous environment.
3Ease of operation
If diverse data types from different edge devices are processed independently, then device autonomy is maintained, but the system cannot leverage synergies across different data types for improved machine learning
Solution Approach 1:
The system implements multi-functionality by enabling leader devices to handle multiple data types from diverse sources within their clusters. Each leader device can process various data types (sensor data, image data, video data) and facilitate cross-data-type model training, allowing devices to maintain autonomy while leveraging synergies across different data types.
Solution Approach 2:
The system uses feedback mechanisms where leader devices share model performance results and updated models across the network. This feedback loop allows autonomous devices to improve their local models based on collective learning from diverse data types, enhancing machine learning effectiveness while preserving device autonomy.
Data Source
AI summary
A machine learning heterogeneous edge device, method, and system are disclosed. In an example embodiment, an edge device includes a communication module, a data collection device, a memory, a machine learning module, a group determination module, and a leader election module. The edge device analyzes collected data with a model, outputs a result, and updates the model to create a local model. The edge device communicates with other edge devices in a heterogeneous group. The edge device determines group membership and determines a leader edge device. The edge device receives a request for the local model, transmits the local model to the leader edge device, receives a mixed model created by the leader edge device performing a mix operation of the local model and a different local model, and replaces the local model with the mixed model.