Partial Information Sharing Over Social Learning Networks

This work addresses the problem of sharing partial information within social learning strategies. In learning, agents solve a distributed multiple hypothesis testing by performing two operations at each instant: first, incorporate from private observations to form their beliefs over set hypotheses; second, combine entirety locally among neighbors. Within sufficiently informative environment and as long connectivity network allows diffuse across agents, these algorithms enable learn true hypothesis. Instead beliefs, this considers case in which will only share regarding one interest, with purpose evaluating its validity, draws conditions under policy does not affect truth learning. We propose approaches for information, depending on whether behave self-aware manner or not. The results show how different regimes arise, approach employed inherent characteristics inference problem. Furthermore, analysis interestingly points possibility deceiving network, evaluated interest is close enough truth.