Distributed Online Learning in Social Recommender Systems

Abstract

Abstract-In this paper, we consider decentralized sequential decision making in distributed online recommender systems, where items are recommended to users based on their search query as well as their specific background including history of bought items, gender and age, all of which comprise the context information of the user. In contrast to centralized recommender systems, in which there is a single centralized seller who has access to the complete inventory of items as well as the complete record of sales and user information, in decentralized recommender systems each seller/learner only has access to the inventory of items and user information for its own products and not the products and user information of other sellers, but can get commission if it sells an item of another seller. Therefore the sellers must distributedly find out for an incoming user which items to recommend (from the set of own items or items of another seller), in order to maximize the revenue from own sales and commissions. We formulate this problem as a cooperative contextual bandit problem, analytically bound the performance of the sellers compared to the best recommendation strategy given the complete realization of user arrivals and the inventory of items, as well as the context-dependent purchase probabilities of each item, and verify our results via numerical examples on a distributed data set adapted based on Amazon data. We evaluate the dependence of the performance of a seller on the inventory of items the seller has, the number of connections it has with the other sellers, and the commissions which the seller gets by selling items of other sellers to its users.Index Terms-Multi-agent online learning, collaborative learning, distributed recommender systems, contextual bandits, regret.

Context-Aware Hierarchical Online Learning for Performance Maximization in Mobile Crowdsourcing

“…For example, the contextual zooming algorithm [29] proposes a non-uniform adaptive partition of the context space. Moreover, [30], [31] use uniform and non-uniform adaptive partitions of the context space. In [32], [33], these algorithms are applied to a wireless communication scenario.…”

Differentially Private Online Learning for Cloud-Based Video Recommendation With Multimedia Big Data in Social Networks

“…This is typically accomplished by adding noise to the item covariance matrix, to hide small changes that arise from a single users contribution. Ashwin et al [8] and [19] CB \ None C [20], [21] CF \ None C [22], [23] GB \ None C [25] CA \ None C [32] CA \ None D [11] Hybrid CF A User C [27] CF Cr User C [13] GB DP User D [14] CF DP User C [8] GB DP User C Our work CA DP User, service provider D Jorgensen [13] combine differential privacy with social graph for recommendation. But their work only study the privacy of sensitive user-item preferences and connections between people, rather than individual features.…”

“…There exist some works studying the contextual bandit [29], [30], where the best action given the context is learned online. C. Tekin et al first proposed a distributed contextual bandit framework for big data classification [31] and social recommendations [32]. But the uniform partition method proposed in their work does not fit into the sparse big data.…”

“…Actually, most OSNs recommend video content to their users based on the user's rich context information (e.g., social status, ages, professions, health conditions and hobbies) contained in their released multimedia data. Regarding this way, several recommendation systems have been proposed [32], [24].…”

Context-Aware Online Spatiotemporal Traffic Prediction

“…However, most of them provide algorithms which are asymptotically converging to an optimal or locally-optimal solution without providing any rates of convergence. On the contrary, we prove regret bounds that hold uniformly over time; the proving technique is adapted from contextual multiarmed bandits (MAB) framework [14]. Since the learner can observe the rewards of all classifiers, the considered problem is more related to prediction with expert advice (PWEA) [15].…”