Probabilistic quantization of unbiased broadcast gossip algorithms for consensus in distributed networks

Abstract

In this paper, probabilistic quantization on Unbiased Broadcast Gossip Algorithms (UBGA) is introduced to adapt to the limited bandwidth of distributed network channels. This innovative quantized consensus algorithm is called Quantized UBGA (QUBGA). By briefly describing the graph and time model, and reviewing UBGA, it is proved that QUBGA can converge in expectation and in the second moment for strongly connected digraphs. Furthermore, it is proved that the convergence rate of QUBGA is related to quantized resolution and topology structure. Performance analysis results are consistent with theoretical analysis.