Separable Multi-Stage Data Processing Model for Edge-Cloud Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to deploy large data processing models on low-capacity devices due to limited computing resources, resulting in long processing times and unsatisfactory accuracy even after model compression.
Innovation Solution
The proposed solution involves using separate two-stage data processing models, where a first model with fewer computing resources processes data on a low-capacity device to generate an initial result with lower accuracy but smaller data size, which is then adjusted by a second model with more computing resources on a higher-capacity device to achieve higher accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If a large data processing model is deployed on a low-capacity device, then the processing capability is improved, but the device cannot handle it due to limited computing resources
Solution Approach 1:
The patent divides a large data processing model into multiple smaller sub-models that can be executed sequentially on low-capacity devices. This segmentation allows the system to achieve the processing capability of a large model while using only the limited computing resources available on edge devices, as each sub-model is small enough to be handled by the device.
2Device complexity
If model compression is applied to reduce computing resource usage, then the device compatibility is improved, but the accuracy of the model deteriorates
Solution Approach 1:
Instead of compressing a single large model which loses accuracy, the patent segments the model into multiple sub-models. Each sub-model maintains sufficient accuracy for its specific function while being small enough to run on low-capacity devices. The sequential execution of these sub-models preserves the overall processing accuracy that would otherwise be lost in compression.
Solution Approach 2:
The patent transforms the problem from a single-dimension trade-off (model size vs. accuracy) to a multi-dimensional solution by distributing the model across multiple sub-models executed in sequence. This dimensional change allows the system to achieve both small model size and high accuracy simultaneously, as each sub-model operates independently with optimized parameters for its specific task.
3Measurement precision
If a single large model is used to maintain high accuracy, then the processing precision is improved, but the processing time increases due to limited device capacity
Solution Approach 1:
The patent segments a large accurate model into multiple smaller sub-models that can be executed quickly on low-capacity devices. While each sub-model processes only a portion of the data, their sequential execution achieves the same overall accuracy as the original large model while significantly reducing processing time, as each sub-model completes its task much faster than the large model would process the entire data.
Data Source
AI summary
Embodiments of the present disclosure provide a method, an electronic device, and a computer program product for data processing. In a method for data processing, a first electronic device processes data based on a first data processing model to generate an initial result. A data size of the initial result is smaller than a data size of the data. The first electronic device sends the initial result to a second electronic device. The initial result is adjusted at the second electronic device and based on a second data processing model to generate an adjusted result. The second electronic device has more computing resources than the first electronic device, the second data processing model occupies more computing resources than the first data processing model, and an accuracy of the adjusted result is higher than that of the initial result.