Transient analysis of multitask learning over adaptive networks with wireless links

Abstract

Distributed multitask learning over adaptive networks with non-ideal links is studied in this paper. The performance of an adaptive multitask network with wireless communication links suffering from flat fading and additive white Gaussian noise is derived through simulation and theoretical analysis. The effect of the wireless link on the learning performance of a multitask network is analyzed. It is shown that this destroying effect is reduced by using local equalizers at each agent along with the intelligent cooperation policy based on the correntropy criterion. Stability condition, steady-state and transient performances are studied theoretically. Finally, the theoretical results are validated by numerical simulations and computer experiments.