Online influence maximization in the absence of network structure

Abstract

In online influence maximization, a learner aims to find a specified number of nodes that have the greatest influence in a network, by iteratively selecting seed nodes (i.e., initially activated nodes) and updating its knowledge of the network via activation feedback. Existing approaches to this problem customarily assume that the structure of the network is known in advance, and focus on how to utilize activation feedback to reveal the features of seed nodes in each iteration, regardless of non-seed nodes which occupy the majority of the node set. In this paper, we present a novel learning framework to carry out online influence maximization in the absence of network structure. In our framework, the underlying influence relationships between nodes are inferred based on activation feedback, and then the influence reachabilities of both seed nodes and non-seed nodes are updated with the latest inferred influence relationships, so that more knowledge about the network can be used to guide the selection of seed nodes in the next iteration. Extensive experiments on both synthetic and real-world networks are conducted, and the results verify the efficacy of our proposed framework.