Split Prediction System for IoT Model Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
IoT devices face challenges in maintaining prediction accuracy due to limited computational power and changing data conditions, leading to reduced reliability of local models compared to larger models running on more powerful servers.
Innovation Solution
Implementing a split prediction system where IoT devices (hub or edge devices) process data locally and transmit it to a provider network for correction using more accurate models, allowing for periodic updates of local models based on new data and leveraging model tiering across multiple devices for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a simple data processing model is implemented on an IoT device, then the device can perform local data analysis and generate predictions, but the reliability and accuracy of predictions deteriorate compared to larger models on powerful computing devices
Solution Approach 1:
The system segments the data processing function into two parts: a simple local model on the IoT device for immediate processing and a larger remote model on the provider network for accurate predictions. The local model handles real-time data analysis while the remote model provides periodic corrections and updates, resolving the contradiction between local capability and prediction accuracy.
Solution Approach 2:
The system implements feedback by having the IoT device send data to the provider network, which returns corrections and updates to the local model. This feedback loop allows the simple local model to maintain and improve its accuracy over time by learning from the more powerful remote model, thereby resolving the reliability issue.
2Adaptability or versatility
If a local model is trained on initial data, then the device can operate independently, but the model loses accuracy over time when data conditions change
Solution Approach 1:
The system makes the local model dynamic by enabling it to receive periodic updates and corrections from the provider network. Instead of being static after initial training, the model adapts to changing data conditions through continuous feedback, resolving the contradiction between initial adaptability and long-term reliability.
Solution Approach 2:
The provider network performs preliminary actions by training and preparing updated models offline, then deploying them to IoT devices. This allows the local models to be proactively updated with new training data and corrections before accuracy degradation becomes problematic, maintaining reliability over time.
3Reliability
If more parameters are used in a model, then prediction accuracy improves, but the computational requirements exceed the capabilities of IoT devices
Solution Approach 1:
The system uses copying by deploying a simplified version of the large provider network model to the IoT device. The local model is a lightweight copy that performs essential functions, while the full-accuracy model remains on the provider network. This resolves the contradiction by providing accurate predictions through the remote model while maintaining device simplicity.
Data Source
AI summary
A model training service of a provider network receives data from edge devices of a remote network. The model training service analyzes the received data. The model training service may also analyze global data from other edge devices of other remote networks. The model training service may then generate updates to local data processing models based on the analysis. The updates are configured to update the local data processing models at the edge devices of the remote network. The provider network deploys the updates to the remote network. The updates are then applied to the data processing models of the edge devices.