Paper Link: https://caod.oriprobe.com/articles/58047775/Sequence_To_Sequence_Learning_for_Online_Imputation_of_Sensory_Data.htm

DOI: 10.15878/j.cnki.instrumentation.2019.02.008

Missing data, in general, could cause huge problems to many applications, such as online sensing in IoT networks, data-driven decision making, both pre- and post-processing in data pipeline, etc.

Traditionally, there are quite a few statistical methods to solve missing data issues. But they usually have many limitations. In this published work, we propose a Seq2Seq model that is based on RNN encoder-decoder architecture to solve data imputation problems.

Below is an overview of our proposed neural network architecture.