Distributed Evidential EM Algorithm for Classification in Networks with Data with Uncertainty
Fang Liu, Kornel Medvenko, Roberto Fox · 2021
In this paper, the issue of data classification in distributed sensor networks with measurements with uncertainty has been investigated. Measurement of sensors in sensor networks can be provided using a Gaussian hybrid model. In this paper, the data are first generated by a combination of Gaussian components and then uncertainty is added to them. Then, a new distributed algorithm called Evidential EM algorithm is used to estimate Gaussian components in the hybrid model. The proposed algorithm is actually an extended version of the EM algorithm for estimating and classifying uncertain data, which consists of two parts of averaging and maximization. Finally, the performance of the proposed algorithm is shown by a simulation example.