Edge Device Data Filtering for Cloud Model Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge-cloud computing systems face challenges in efficiently reducing data transfer from the edge side to the cloud for learning inference models, leading to network bandwidth oppression and inefficient data processing, as they lack the ability to selectively transfer data based on its relevance and usefulness.
Innovation Solution
A learning system comprising a cloud computing system and an edge device with sensors, where the edge device transfers detection data only when it is related to the inference model or when the model's certainty factor is low, and uses a rarity evaluation to classify data for learning likelihood, thereby reducing unnecessary data transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all detection data is transferred from edge to cloud for learning inference models, then the learning model can be trained with comprehensive data, but network bandwidth is oppressed and data transfer efficiency is reduced
Solution Approach 1:
The patent extracts only the necessary data for model training by implementing a filtering mechanism at the edge device. The edge device determines whether detection data should be transferred to the cloud based on pre-stored inference models and data relevance criteria, thereby extracting only useful data while discarding redundant information before transmission.
Solution Approach 2:
The patent applies preliminary action by pre-loading inference models to the edge device before actual data collection. This allows the edge device to evaluate detection data against existing models in advance and make informed decisions about data transfer, avoiding unnecessary bandwidth consumption while ensuring training quality.
2Loss of energy
If detection data is selectively transferred based on relevance, then network bandwidth is saved, but the edge device must perform complex evaluation decisions
Solution Approach 1:
The patent reduces edge device complexity by performing the heavy lifting of model evaluation in advance. Inference models are pre-loaded to the edge device, enabling it to make rapid relevance decisions without complex real-time computation, thus saving bandwidth while keeping edge processing simple.
Solution Approach 2:
The patent uses copying by replicating inference models at the edge device. Instead of transferring all raw data to the cloud for processing, the system copies the model to the edge, allowing local evaluation of data relevance. This shifts the computational burden from the edge to the cloud while simplifying edge device operations.
3Productivity
If only data related to current inference model is transferred, then data transfer is optimized, but new insights and model improvements may be missed
Solution Approach 1:
The patent implements feedback by continuously transferring detection data to the cloud for analysis, even if it doesn't immediately match current inference models. The cloud side analyzes this data and provides feedback by generating updated models or adjusting existing models, ensuring the system adapts to new patterns and improves over time.
Solution Approach 2:
The patent applies dynamics by making the data transfer strategy adaptive rather than static. The system dynamically adjusts what data is transferred based on current models, new discoveries, and changing conditions. Inference models are continuously updated based on analyzed data, allowing the system to adapt to new insights while maintaining transfer efficiency.
Data Source
AI summary
A system includes a cloud computing system having a server, and a user environment computing system having an edge electronic device and a group of edge side sensors installed on at least one of inside and outside of the edge electronic device, the cloud computing system and the user environment computing system connected via a network line. The user environment computing system transfers a plurality of kinds of detection data collected by the group of edge side sensors to the cloud computing system for learning of an inference model generated by the server. When detection data is newly obtained from one sensor out of the group of edge side sensors and the detection data newly obtained is related to the inference model, the detection data newly obtained is transferred to the server for additional learning of the inference model.