A Distributed Incremental Learning Method for Sparse Kernel Machine over Wireless Sensor Network

In wireless sensor networks, centralized learning methods have very high communication costs and energy consumption. These are caused by the need to transmit scattered training samples from various sensor nodes to the central fusion center where a classifier or a regression machine is trained by batch learning with all training samples. To decrease the communication costs and energy consumption, a distributed learning problem of kernel machine by incorporating e1 norm regularization is investigated, and a novel distributed incremental learning algorithm for the e1-regularized kernel minimum mean squared error (KMSE) machine is proposed. The proposed algorithm relies on the incremental learning method and the collaboration between single-hop neighboring nodes based on Markov Chains Theory. This paper evaluates the proposed algorithm with respect to the prediction accuracy, the sparse rate of model and the communication costs on synthetic and real datasets. The simulation results show that the proposed algorithm can obtain approximately the same prediction accuracy as that obtained by the batch learning method and the current state-of-the-art distributed algorithms. Moreover, it is significantly superior in terms of the sparse rate of model and communication costs.