Learning the Truth in Social Networks Using Multi-Armed Bandit

Abstract

This paper explains how agents in a social network can learn the arbitrary time-varying true state of the network. This is practical in social networks where information is released and updated without any coordination. Most existing literature for learning the true state using the non-Bayesian learning approach, assumes that this true state is fixed, which is impractical. To address this problem, the social network is modeled as a graph network, and the time-varying true state is treated as a multi-armed bandit problem. The few works that have applied multi-armed bandit to a social network did not take into consideration the adversarial effects. Therefore, this paper proposes two non-stochastic multi-armed bandit algorithms that can handle the time-varying true state, even in the presence of an oblivious adversary. Regret bounds on the algorithms are obtained, and the simulation performance shows that all agents can converge to the most stable state. The sublinearity of the proposed algorithms is also compared with two well-known non-stochastic multi-armed bandit algorithms.INDEX TERMS strongly connected network, non-Bayesian learning, diffusion learning, multi-armed bandit, regret.

Differential Privacy in Social Networks Using Multi-Armed Bandit

“…The speed of convergence for Algorithm 1 is nearly 23% slower on average when compared to the speed of convergence of the non-private algorithm in [16]. However, algorithm 2 is over 56% slower on average, than the nonprivate algorithm.…”

“…∀θ ∈ End and any spying neighboring agents seeking to know more information about the agent than what is necessary. The non-private algorithm of these proposed algorithms is found in [16]. The non-private version leaks information to a thirdparty intruder and a spying agent.…”

“…The definition of regret in (10) is slightly different from the definition of regret in [16]. The p k,t (θ ) in the definition of the regret in [16] is replaced with µ k,t (θ ) in (10). Hence, we will compute the regret bound.…”

“…Also, information is arbitrarily released and updated in social networks. Recent studies show that incorporating non-stochastic multi-armed bandit techniques -rather than stochastic multi-armed bandit techniques -into non-Bayesian learning approaches can effectively track this time-varying true state [14]- [16]. Non-stochastic multi-armed bandit is a variant of the online learning strategies that work well in sequential decision-making.…”

Online Learning of Time-Varying Unbalanced Networks in Non-Convex Environments: A Multi-Armed Bandit Approach

“…In such cases, no agent fully learns about the network; hence, there is a need for cooperation among neighboring agents to acquire global knowledge of the network. Some practical applications of distributed networks are decentralized tracking, estimation and detection, social networks, sensor networks, smart grids, and many optimization problems [2], [3], [4], [5], [6], [7], [8], [9].…”