.png){kind=icon}
.png){kind=icon}
Return
Lie Group feature compressed sensing joint Encoder-Decoder network for data classification and anomaly detection
.png){kind=icon}
PRE
AI
DOI：10.1016/j.adhoc.2026.104384.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A universal classification module of compressed sensing (CS) based on the feature of Lie Group intrinsic mean is designed. The Lie Group intrinsic mean feature compressed sensing technique is introduced to characterize different data, and which sensors (features) are more relevant in the whole sensor set are comprehensively analyzed. • The architecture of Encoder-Decoder is improved. Considering the diverse scenes and diverse types of sensor node data, the classified data is set as the input of the improved Encoder-Decoder module. • Extensive experiments on a large number of different real-world data sets verify that the classification accuracy and detection performance of our proposed approach is superior to other approaches, which also shows that our approach has strong feature representation, classification, and detection capabilities. In addition, the influence of block size and adversarially learned on performance are explored.
Keywords:
Artificial intelligence application
Anomaly detection
Compressed sensing
Data classification
Internet of Things
Lie group machine learning