Influence Maximization aims to select a subset of elements in a social network to maximize information spread under a diffusion model. While existing work primarily focuses on selecting influential nodes, these approaches assume unrestricted message propagation-an assumption that fails in closed social networks, where content visibility is constrained and node-level activations may be infeasible. Motivated by the growing adoption of privacy-focused platforms such as Signal, Discord, Instagram, and Slack, our work addresses the following fundamental question: How can we learn effective edge activation strategies for influence maximization in closed networks? To answer this question we introduce Reinforcement Learning for Link Activation (RELINK), the first DRL framework for edge-level influence maximization in privacy-constrained networks. It models edge selection as a Markov Decision Process, where the agent learns to activate edges under budget constraints. Unlike prior node-based DRL methods, RELINK uses an edge-centric Q-learning approach that accounts for structural constraints and constrained information propagation. Our framework combines a rich node embedding pipeline with an edge-aware aggregation module. The agent is trained using an n-step Double DQN objective, guided by dense reward signals that capture marginal gains in influence spread. Extensive experiments on real-world networks show that RELINK consistently outperforms existing edge-based methods, achieving up to 15% higher influence spread and improved scalability across diverse settings.

In contemporary society, groups are pivotal in shaping decisions and actions. The consensus of a majority of members on specific topics often guides the collective decision-making in groups. Group influence maximization (GIM) aims to select$k$seed users in a network to maximize the number of eventually activated groups. A group is said to be activated if$\beta$percent of users in this group are activated. This study delves into the strategic selection of seed users in social networks to maximize the spread of a topic, thereby activating the highest number of groups. The GIM problem, inherently NP-hard when computing the influence spread from a selected set of nodes, has traditionally faced obstacles in ensuring theoretical robustness, time efficiency, and adaptability in large and complex network environments. To overcome these challenges, we introduce a robust framework called GIMDRL that addresses the GIM problem in social networks using deep reinforcement learning (DRL). Our approach integrates node embeddings from multiple graph neural networks, thereby utilizing diverse information for effective network analysis. This integration plays a crucial role in optimizing the parameter learning process. Extensive experiments are conducted on real-world and synthetic datasets to assess the performance of our proposed framework. The results of these experiments indicate that our approach significantly outperforms existing methods in GIM, even when trained on sampled graphs. This highlights our model's strong capacity for generalization in varying network scenarios.

Existing solutions for rumor containment typically model social networks as regular graphs, focusing solely on dyadic relationships between pairs of individuals. However, these solutions overlook the crowd influence, which arises from higher order relationships involving multiple individuals. This crowd influence is a distinct concept that cannot be substituted by the cumulative effect of individual influences through dyadic relationships. Therefore, it is important to consider the impact of crowd influence when modeling influence diffusion in the network. Moreover, traditional rumor containment methods lack generalization capacity. These methods require complete reexecution of algorithms whenever the target network changes, rendering them less efficient. In this work, we model the network as a hypergraph to effectively capture the crowd influence through higher order social relationships. We propose RCDRL-H, adeep reinforcement learningframework capable of constructing a trained model for containing rumors in previously unseen networks. Additionally, we introducehyper-structure2vec, a node embedding technique for hypergraphs, and an efficient rumor containment estimation function that eliminates the need for computationally expensive Monte Carlo simulations during training. Experiments carried out on real-world social networks demonstrate that improved rumor containment can be achieved by treating the network as a hypergraph. It has also been observed that the achieved rumor containment is close to that of the greedy approach. Moreover, using the new estimation function significantly reduces training time.

Social networks are becoming important dissemination platforms, and a large body of works have been performed on viral marketing, but most are to maximize the benefits associated with the number of active nodes. In this paper, we study the benefits related to interactions among activated nodes. Furthermore, due to the uncertainty in edge probability estimates in social networks, we propose the robust profit maximization problem to have the best solution in the worst case of probability settings. We design a double sandwich algorithm to this problem and further improve the algorithm with sampling method such that it increases robustness of the output. Through real data sets, we verify the effectiveness of our proposed algorithm.

Group plays an important role in social society. Much of the world's decision or work is done by groups and teams. A group's decision should be made based on most of the members in the group that reach agreement on a concerned topic. If we want to spread a topic and maximize the total number of activated groups in a social network, which seed users should we choose. In this article, we will study a new influence maximization (IM) problem which focuses on the number of groups activated by some concerned topic or information. A group is said to be activated if β percent of users in this group are activated. Group IM (GIM) aims to select k seed users such that the number of eventually activated groups is maximized. We first analyze the complexity and approximability of GIM, which is NP-hard, and the objective function presented in this article is proven to be neither submodular nor supermodular. We develop an upper bound problem and a lower bound problem whose objective functions are submodular. Then, an algorithm based on group coverage will be proposed, and the Sandwich framework is formulated with theoretical analysis to solve GIM. Our experiments verify the effectiveness of our method, as well as the advantage of our method against the other heuristic methods.