Sensor Node Autoencoder Split for Low-Bandwidth Classifier Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In sensor networks, updating a classifier's classification label is challenging due to the difficulty in collecting and processing learning data at sensor nodes, which lack sufficient storage capacity, processing power, and power for large-scale data learning, leading to inefficient communication and varying classifier performance across nodes.
Innovation Solution
Implementing an autoencoder with an encoder at the sensor node to compress data and a decoder at the data center to restore it, allowing for reduced communication volume and enabling the collection of learning data for updating classification labels, using techniques like support vector machines or neural networks for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If sensor data is directly transmitted to data center, then communication simplicity is maintained, but communication volume exceeds bandwidth limit
Solution Approach 1:
The patent segments the data processing function by dividing the autoencoder into two parts: the encoder at the sensor node and the decoder at the data center. This segmentation allows local preprocessing (encoding) to reduce communication volume while maintaining the ability to reconstruct full data when needed, thus resolving the contradiction between communication simplicity and volume control.
2Adaptability or versatility
If classification label is changed in course of operation, then adaptability is improved, but learning data collection becomes difficult at sensor node
Solution Approach 1:
The patent solves the learning data collection problem by transmitting encoded data (compressed representation) from sensor nodes to the data center, where decoding and learning data pairing occur. This dimensional transformation allows the system to maintain adaptability for changing classification labels while avoiding the complexity of collecting and storing raw sensor data at resource-constrained sensor nodes.
3Quantity of substance
If sensor node performs local processing, then communication volume is reduced, but processing power and storage capacity are insufficient
Solution Approach 1:
The patent segments the computationally intensive autoencoder functionality across different locations: the encoder (lighter computational burden) resides at the sensor node to reduce communication volume, while the decoder (heavier computational burden) and learning data storage reside at the data center with sufficient processing power and storage capacity.
4Adaptability or versatility
If learning data is stored at data center, then learning capability is improved, but communication volume increases when transmitting sensor data
Solution Approach 1:
The patent changes the parameter of data representation by encoding sensor data into a compressed form before transmission. This parameter transformation (from raw sensor data to encoded data) reduces communication volume while preserving the essential information needed for learning, allowing the data center to maintain learning capability without receiving full-volume raw data.
Data Source
AI summary
Provided is a sensor network capable of acquiring, from a data center, learning data composed of a pair of sensor data (input) and a classification label (output) necessary for adding/changing a classification label for updating a classifier while reducing the volume of communication between a sensor and the center. One aspect of the present invention relates to a sensor network, including one or more sensor nodes and a data center. The sensor node includes: an encoding unit for encoding sensor data by an encoder part of an autoencoder; and a transmission unit for transmitting the encoded data. The data center includes: a reception unit for receiving data encoded from sensor data by the encoder part of the autoencoder; a decoding unit for decoding the encoded data by a decoder part of the autoencoder; and a storage unit for storing the decoded data therein.