Simplified Machine Learning Model Deployment on Edge Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models, such as neural networks, are computationally expensive and power-intensive, making them unsuitable for deployment on resource-constrained edge devices, and there are limited methods for updating these models after initial deployment to keep them current.
Innovation Solution
A computer-implemented method for updating a simplified representation of a machine learning model by re-training it based on data from edge devices and generating a simplified representation of the re-trained model for deployment back to the edge device, allowing for efficient execution and lifecycle management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional machine learning models are deployed on edge devices, then model accuracy and performance are improved, but computational resource consumption and power usage increase significantly
Solution Approach 1:
The patent segments the machine learning model into a simplified representation that can be deployed on edge devices. This segmentation allows the model to maintain sufficient accuracy for edge computing while reducing computational complexity and power consumption requirements.
Solution Approach 2:
The patent transforms the conventional machine learning model by changing its parameters and structure to create a simplified representation. This parameter transformation reduces the model's computational footprint while preserving essential predictive capabilities for deployment on resource-constrained edge devices.
2Reliability
If conventional machine learning models are deployed on edge devices, then model accuracy is improved, but device complexity increases
Solution Approach 1:
The model is segmented into a simplified representation with reduced architectural complexity. This segmentation maintains core accuracy functions while eliminating unnecessary computational layers, making the model suitable for edge devices with limited processing capabilities.
Solution Approach 2:
The patent creates a simplified, lightweight version of the machine learning model that sacrifices some computational complexity for deployability on edge devices. This simplified representation acts as a practical, resource-efficient alternative to the full conventional model.
3Speed
If machine learning models are deployed on edge devices, then real-time processing capability is improved, but model updating capability deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where execution data from the simplified model on edge devices is collected and used to retrain and update the model. This feedback loop enables continuous model improvement and adaptation while maintaining real-time processing capabilities on resource-constrained devices.
Solution Approach 2:
The system performs preliminary actions by collecting execution data during model operation and using this data to retrain updated versions of the simplified model. This preliminary data collection and model updating process enables the system to adapt to new patterns while maintaining real-time processing performance.
Data Source
AI summary
One embodiment of a method for updating a simplified representation of a machine learning model includes receiving, from an edge device, data associated with execution of the simplified representation of the machine learning model on the edge device, performing one or more operations to re-train the machine learning model based on at least a portion of the data to generate a re-trained machine learning model, generating a simplified representation of the re-trained machine learning model, and transmitting, to the edge device, the simplified representation of the re-trained machine learning model for execution on the edge device.