PROJECT TITLE :

FINE: A Framework for Distributed Learning on Incomplete Observations for Heterogeneous Crowdsensing Networks - 2018

ABSTRACT:

Lately, there has been a wide selection of applications of crowdsensing in mobile social networks and vehicle networks. As centralized learning ways cause unreliabitlity of information assortment, high cost of central server, and concern of privacy, one vital drawback is how to hold out an accurate distributed learning method to estimate parameters of an unknown model in crowdsensing. Motivated by this, we tend to present the design, analysis, and analysis of FINE, a distributed learning framework for incomplete-knowledge and non-sleek estimation. Our design, dedicated to develop a feasible framework that efficiently and accurately learns the parameters in crowdsensing networks, well generalizes the previous learning ways in that it supports heterogeneous dimensions of data records observed by totally different nodes, along with minimization based mostly on non-sleek error functions. In particular, FINE uses a unique distributed record completion algorithm that allows each node to get the worldwide consensus by an economical communication with neighbors, and a distributed dual average algorithm that achieves the efficiency of minimizing non-swish error functions. Our analysis shows that every one these algorithms converge, of which the convergence rates are derived to substantiate their efficiency. We evaluate the performance of our framework with experiments on artificial and real-world networks.