Distributed Sensor Data Prioritization for Centralized ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning model training methods, such as centralized and federated learning, face challenges in battery life and data transfer efficiency, particularly in resource-constrained environments, leading to reduced model accuracy and limited adoption.
Innovation Solution
A method and system for optimizing sensor data collection by identifying a data prioritization technique, prioritizing data based on similarity and system constraints, and transmitting it to a centralized repository for training, using an autoencoder to evaluate and adjust the technique.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is transmitted from edge devices to the cloud for centralized training, then model accuracy is improved, but battery life is reduced due to significant current consumption from Bluetooth or WiFi signals
Solution Approach 1:
The patent applies preliminary action by performing data prioritization and selection at the edge device before transmission. The system pre-processes sensor data to identify and prioritize under-represented operational states, creating an optimized data subset that maintains model training effectiveness while reducing the volume of data requiring transmission, thereby conserving battery energy.
Solution Approach 2:
The patent implements local quality by applying different processing treatments to different portions of sensor data based on their operational state characteristics. Instead of uniform data collection, the system selectively prioritizes data from under-represented operational states while potentially reducing or eliminating transmission of data from well-represented states, creating non-uniform data transmission quality that optimizes both model accuracy and energy consumption.
2Measurement precision
If all sensor data is transmitted to the centralized repository, then model training accuracy is improved, but data transfer efficiency and network bandwidth utilization are reduced
Solution Approach 1:
The patent applies the extraction principle by removing unnecessary data from the transmission set. The data prioritization technique extracts and identifies under-represented operational states, then selectively transmits only this prioritized subset to the centralized repository, excluding redundant data from well-represented operational states, thereby improving data transfer efficiency while maintaining training accuracy.
Solution Approach 2:
The system performs preliminary data analysis and prioritization at the edge device before transmission, identifying under-represented operational states in advance. This pre-processing step creates an optimized data subset that eliminates the need to transmit all raw sensor data, thereby improving network bandwidth utilization and data transfer efficiency.
3Duration of action of moving object
If data prioritization is implemented to reduce transmission volume, then battery life and data transfer efficiency are improved, but data selection complexity increases
Solution Approach 1:
The patent introduces an intermediary mechanism in the form of automated data prioritization algorithms and evaluation frameworks that bridge the gap between simple data collection and complex selective transmission. The system uses intermediate data structures, priority queues, and automated evaluation metrics to manage the selection process, reducing the perceived complexity for users while maintaining sophisticated data selection capabilities.
4Duration of action of moving object
If federated learning is used to train models locally, then battery life is improved by avoiding data transmission, but model accuracy deteriorates due to differences in local data distributions among agents
Solution Approach 1:
The patent applies preliminary action by performing data prioritization and selection at the edge device before transmission. The system pre-processes sensor data to identify and prioritize under-represented operational states, creating an optimized data subset that maintains model training effectiveness while reducing the volume of data requiring transmission, thereby conserving battery energy.
Solution Approach 2:
The patent implements local quality by applying different processing treatments to different portions of sensor data based on their operational state characteristics. Instead of uniform data collection, the system selectively prioritizes data from under-represented operational states while potentially reducing or eliminating transmission of data from well-represented states, creating non-uniform data transmission quality that optimizes both model accuracy and energy consumption.
Data Source
AI summary
Some embodiments relate to a method and system of optimizing sensor data collection for centralized training of machine learning models. The method comprises identifying a data prioritization technique; collecting sensor data and implementing the identified data prioritization technique to prioritize the data for storage and transmission to a centralized repository; transmitting the data to the centralized repository; training the machine learning model on the centralized repository using the transmitted data to predict or identify events, such as failure in vehicles and other machinery; evaluating the data prioritization technique using a forward pass of an autoencoder to create an output; and determining an adjustment to the data prioritization technique based on the output.