Data Aided Sensing with Variational Autoencoder

In this paper, we propose an intelligent data-aided sensing method exploiting variational auto-encoder (VAE) for wireless sensor networks. A set of measurements of all sensors is analyzed to identify the underlying cause. For this purpose, we employ
$\beta$
- VAE to find the cause with a partial observation of the measurement vector and utilize the inferred cause for communication efficient data collection in uploading the data. The resulting approach can be used when collecting all measurements would be expensive, or excessively time consuming.