Edge Device ML Model Updates via Markup Documents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Edge devices face limitations in processing and storage capacity, making it infeasible to implement machine learning exclusively on these devices, and sharing large volumes of data for global machine learning is impractical due to network resource constraints, leading to suboptimal performance and localized results.
Innovation Solution
A system where a centralized server computer system trains a global machine learning model using data from multiple edge devices and sends model updates to edge devices via markup language documents, enabling edge devices to use a global model for improved local processing and reducing the need for continuous network connectivity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If machine learning is implemented exclusively on edge devices, then local processing capability is improved, but device complexity and computational requirements exceed edge device limitations
Solution Approach 1:
The machine learning system is segmented into two parts: a global model trained on a centralized server using large datasets, and a local model deployed on edge devices for inference. This segmentation allows complex computational tasks to be performed on the server while simpler tasks run on resource-constrained edge devices, resolving the contradiction between local processing capability and device complexity limitations.
2Measurement precision
If large volumes of data are shared for global machine learning, then model training quality is improved, but network resource consumption increases
Solution Approach 1:
The invention extracts and transmits only the essential training data or model updates from the centralized server to edge devices, rather than sharing all raw data volumes. This selective extraction maintains model training quality while significantly reducing network resource consumption by transmitting only the necessary information for global model improvement.
3Adaptability or versatility
If raw data is stored for later analysis, then data analysis flexibility is improved, but storage requirements increase
Solution Approach 1:
The system performs preliminary processing and aggregation of raw data at the edge devices before transmission to the centralized server. This preliminary action reduces the volume of data that needs to be stored and transmitted, while still preserving the essential information needed for flexible later analysis. The edge devices prepare data in advance, enabling both storage efficiency and analytical flexibility.
Data Source
AI summary
Disclosed is a technique that can be performed by a server computer system. The technique can include executing a machine learning process to generate a machine learning model based on global data collected from one or more electronic devices, wherein the machine learning model is described by model data. The technique can further include encapsulating the model data in a markup language document. The technique can further include sending, over a network, the markup language document to at least one electronic device of the one or more electronic devices to cause the at least one electronic device to update a local device machine learning model.