Communication Efficient Distributed Estimation Over Directed Random Graphs
Anit Kumar Sahu, Dušan Jakovetić, Dragana Bajović, Soummya Kar · 2019
Recently, a communication efficient recursive distributed estimator, CREDO, has been proposed, that utilizes increasingly sparse randomized bidirectional communications. CREDO achieves order-optimal O(1/t) mean square error (MSE) rate in the number of per-node processed samples t, and a O(1/C2-ςt) MSE rate in the number of per-node communications, where ζ > 0 is arbitrarily small. In this paper, we present directed CREDO, D - CREDO for short - a distributed recursive estimator that utilizes directed increasingly sparse communications. We show that D - CREDO further dramatically improves communication efficiency, achieving the O(1/Cκt ) communication MSE rate with arbitrarily high exponent κ, while keeping the order-optimal O(1/t) samplewise MSE rate. Numerical examples on real data sets confirm our results.