Neural Information Processing Systems http://nips.cc/

Most previous studies focus on the independent cascade (IC) model under the edge-level feedback.

Influence maximization in social networks plays a vital role in applications such as viral marketing, epidemiology, product recommendation, opinion mining, and counter-terrorism. A common approach identifies seed nodes by first detecting disjoint communities and subsequently selecting representative nodes from these communities. However, whether the quality of detected communities consistently affects the spread of influence under the Independent Cascade model remains unclear. This paper addresses this question by extending a previously proposed disjoint community detection method, termed $α$-Hierarchical Clustering, to the influence maximization problem under the Independent Cascade model. The proposed method is compared with an alternative approach that employs the same seed selection criteria but relies on communities of lower quality obtained through standard Hierarchical Clustering. The former is referred to as Hierarchical Clustering-based Influence Maximization, while the latter, which leverages higher-quality community structures to guide seed selection, is termed $α$-Hierarchical Clustering-based Influence Maximization. Extensive experiments are performed on multiple real-world datasets to assess the effectiveness of both methods. The results demonstrate that higher-quality community structures substantially improve information diffusion under the Independent Cascade model, particularly when the propagation probability is low. These findings underscore the critical importance of community quality in guiding effective seed selection for influence maximization in complex networks.

We show P AC learnability of influence functions for three common influence models, namely, the Linear Threshold (L T), Independent Cascade (IC) and V oter models, and present concrete sample complexity results in each case. Our results for the L T model are based on interesting connections with neural networks; those for the IC model are based an interpretation of the influence function as an expectation over random draw of a subgraph and use covering number arguments; and those for the V oter model are based on a reduction to linear regression. We show these results for the case in which the cascades are only partially observed and we do not see the time steps in which a node has been influenced. We also provide efficient polynomial time learning algorithms for a setting with full observation, i.e.

Most previous studies focus on the independent cascade (IC) model under the edge-level feedback.

Most previous studies focus on the independent cascade (IC) model under the edge-level feedback.

Influence maximization and its variants have applications in viral marketing, rumor control, etc. and

In social networks, people influence each other through social links, which can be represented as propagation among nodes in graphs. Influence minimization (IMIN) is the problem of manipulating the structures of an input graph (e.g., removing edges) to reduce the propagation among nodes. IMIN can represent time-critical real-world applications, such as rumor blocking, but IMIN is theoretically difficult and computationally expensive. Moreover, the discrete nature of IMIN hinders the usage of powerful machine learning techniques, which requires differentiable computation. In this work, we propose DiffIM, a novel method for IMIN with two differentiable schemes for acceleration: (1) surrogate modeling for efficient influence estimation, which avoids time-consuming simulations (e.g., Monte Carlo), and (2) the continuous relaxation of decisions, which avoids the evaluation of individual discrete decisions (e.g., removing an edge). We further propose a third accelerating scheme, gradient-driven selection, that chooses edges instantly based on gradients without optimization (spec., gradient descent iterations) on each test instance. Through extensive experiments on real-world graphs, we show that each proposed scheme significantly improves speed with little (or even no) IMIN performance degradation. Our method is Pareto-optimal (i.e., no baseline is faster and more effective than it) and typically several orders of magnitude (spec., up to 15,160X) faster than the most effective baseline while being more effective.

This study explores the ability of Image Captioning (IC) models to decode masked visual content sourced from diverse datasets. Our findings reveal the IC model's capability to generate captions from masked images, closely resembling the original content. Notably, even in the presence of masks, the model adeptly crafts descriptive textual information that goes beyond what is observable in the original image-generated captions. While the decoding performance of the IC model experiences a decline with an increase in the masked region's area, the model still performs well when important regions of the image are not masked at high coverage.