Information Propagation Research Articles

A unified scalable framework for causal sweeping strategies for Physics-Informed Neural Networks (PINNs) and their temporal decompositions

Physics-informed neural networks (PINNs) as a means of solving partial differential equations (PDE) have garnered much attention in the Computational Science and Engineering (CS&E) world. However, a recent topic of interest is exploring various training (i.e., optimization) challenges – in particular, arriving at poor local minima in the optimization landscape results in a PINN approximation giving an inferior, and sometimes trivial, solution when solving forward time-dependent PDEs with no data. This problem is also found in, and in some sense more difficult, with domain decomposition strategies such as temporal decomposition using XPINNs. We furnish examples and explanations for different training challenges, their cause, and how they relate to information propagation and temporal decomposition. We then propose a new stacked-decomposition method that bridges the gap between time-marching PINNs and XPINNs. We also introduce significant computational speed-ups by using transfer learning concepts to initialize subnetworks in the domain and loss tolerance-based propagation for the subdomains. Finally, we formulate a new time-sweeping collocation point algorithm inspired by the previous PINNs causality literature, which our framework can still describe, and provides a significant computational speed-up via reduced-cost collocation point segmentation. The proposed methods form our unified framework, which overcomes training challenges in PINNs and XPINNs for time-dependent PDEs by respecting the causality in multiple forms and improving scalability by limiting the computation required per optimization iteration. Finally, we provide numerical results for these methods on baseline PDE problems for which unmodified PINNs and XPINNs struggle to train.

DAG: Dual Attention Graph Representation Learning for Node Classification

Transformer-based graph neural networks have accomplished notable achievements by utilizing the self-attention mechanism for message passing in various domains. However, traditional methods overlook the diverse significance of intra-node representations, focusing solely on internode interactions. To overcome this limitation, we propose a DAG (Dual Attention Graph), a novel approach that integrates both intra-node and internode dynamics for node classification tasks. By considering the information exchange process between nodes from dual branches, DAG provides a holistic understanding of information propagation within graphs, enhancing the interpretability of graph-based machine learning applications. The experimental evaluations demonstrate that DAG excels in node classification tasks, outperforming current benchmark models across ten datasets.

A tale of two roles: exploring topic-specific susceptibility and influence in cascade prediction

We propose a new deep learning cascade prediction model CasSIM that can simultaneously achieve two most demanded objectives: popularity prediction and final adopter prediction. Compared to existing methods based on cascade representation, CasSIM simulates information diffusion processes by exploring users’ dual roles in information propagation with three basic factors: users’ susceptibilities, influences and message contents. With effective user profiling, we are the first to capture the topic-specific property of susceptibilities and influences. In addition, the use of graph neural networks allows CasSIM to capture the dynamics of susceptibilities and influences during information diffusion. We evaluate the effectiveness of CasSIM on three real-life datasets and the results show that CasSIM outperforms the state-of-the-art methods in popularity and final adopter prediction.

Optimal test design for reliability demonstration under multi-stage acceptance uncertainties

A reliability demonstration test (RDT) plays a critical role in safeguarding product reliability and making sure it meets the target requirement. When planning an RDT, the test planning parameters are determined before executing the RDT. There is uncertainty associated with the test result and whether the product will be acceptable and released into the market with additional costs resulting from the warranty service or whether a reliability growth process is needed to further improve the product’s reliability. Potentially, such a process could be repeated multiple times depending on how quickly the reliability growth process can improve product reliability. Existing RDT designs primarily consider the cost of RDT itself or over a single demonstration stage before the next possible RDT, and hence fail to fully address the uncertainty of all possible future RDTs and various pathways a product may go through in a multi-stage demonstration process. By focusing on binomial RDT (BRDT) plans based on failure count data, this paper proposes an optimal Bayesian BRDT design framework by explicitly quantifying the multi-stage acceptance uncertainties resulting from current and subsequent BRDTs. It allows the BRDT planning decision to be determined more holistically by anticipating the costs of warranty service and reliability growth along different pathways over multiple stages. A recursive information propagation algorithm is proposed to incorporate the prior belief of product reliability and allow it to evolve and update over multiple stages of BRDT. A case study is presented to illustrate the proposed multi-stage Bayesian BRDT design framework and demonstrate its cost-efficiency compared to existing strategies. A comprehensive sensitivity analysis is also provided to demonstrate the impact of the relative size of different cost components, reliability growth rate, and prior setting on the performance of the proposed method.

High-resolution CBCT assessing a maxillary central incisor with root bifurcation: A case report

Cone beam computed tomography (CBCT) is an imaging modality widely used in endodontics, since it provides three-dimensional images without superposition of structures. The maxillary central incisor usually presents a single root and one root canal, but anatomical variations can happen. The aim of the present paper is to report a high-resolution CBCT evaluation of an upper central incisor with root bifurcation. This paper shows a descriptive and qualitative assessment of high-resolution CBCT images for endodontic purposes. A 39-year-old male patient was referred to a radiologic clinic to acquire CBCT images of the right upper central incisor, as the referring clinician noticed an anatomical difference between the two upper central incisors on a panoramic radiograph and suspected of dens invaginatus. CBCT images were acquired in high resolution (voxel size of 0.08 mm) and a restricted field-of-view (4x4 cm). CBCT volume displayed an anatomical variation of the central incisor, which presented root bifurcation and two root canals, with a normal clinical crown. The present case report shows a rare diagnosis of a maxillary central incisor presenting bifurcated root with two root canals, and to the best of the authors’ knowledge, this is the first case report to present high-resolution CBCT images with disclosed acquisition protocols to diagnose a bifurcated central incisor. The documentation of unusual cases has great didact value and contributes to information propagation. In the present case, high-resolution CBCT allowed a detailed evaluation of both roots and root canals, facilitating the clinician’s diagnostic task.

SHI1I2R competitive information spreading model in online and offline two-layer networks in emergencies

Previous research on information propagation has primarily focused on either online social networks or offline social networks. Additionally, there remains a gap in understanding the propagation of competitive information in emergency scenarios. This paper introduces an SHI1I2R information spreading model to investigate the dynamics of competitive information dissemination and interactions in online and offline two-layer networks during emergencies. To enhance the accuracy of information propagation modeling, we depart from the conventional assumption of constant spreading and decay rates and instead propose nonlinear spreading and interest decay rates within the SHI1I2R model. Employing a microscopic Markov chain approach, we conduct theoretical analysis and derive the spreading threshold of information in two-layer networks. Subsequently, we perform numerical simulations to validate the effectiveness of the SHI1I2R model and explore the factors influencing information propagation. Our findings indicate that information spreads more rapidly and extensively in two-layer networks compared to one-layer networks. Furthermore, the final scale of information dissemination and the speed of propagation follow the order: random network > small-world network > scale-free network > regular network.

Social networks, disinformation and diplomacy: a dynamic model for a current problem

The potential of social networks for the circulation of disinformation as a strategy of diplomacy has been of great interest to the academic community, but the way in which it is propagated and modelled is still in its beginnings. This article aimed to simulate the propagation of disinformation in social networks derived from the diplomacy strategy, based on the elements of the system. The main research question that was opened up was how do the elements of disinformation derived from the social media diplomacy strategy interact to affect a susceptible population? For the design of the simulation model, system dynamics was used as the main technique in the re-search methodology in conjunction with statistical analysis. Five computational simulations were run for the adoption methods of susceptible and uninformed population, misinformation techniques and echo chamber. The model developed found that the diplomacy disinformation agent is able to spread its message efficiently through the bot outreach mechanism and only a part of the susceptible population unsubscribes to the disinformation agent’s account. Significant differences were identified in the absence of paid outreach, bots and trolls in the propagation of information, and in the variation in the timing of disinformation propagation. Consequently, the developed model allows the understanding of the problem of disinformation as a strategy of diplomacy from international rather than local dynamics, as well as the effects of the use of each element in the system.

Knowledge-Aware Dual-Channel Graph Neural Networks For Denoising Recommendation

Abstract Knowledge graph (KG) is introduced as side information into recommender systems, which can alleviate the sparsity and cold start problems in collaborative filtering. Existing studies mainly focus on modeling users’ historical behavior data and KG-based propagation. However, they have the limitation of ignoring noise information during recommendation. We consider that noise exists in two parts (i.e. KG and user-item interaction data). In this paper, we propose Knowledge-aware Dual-Channel Graph Neural Networks (KDGNN) to improve the recommendation performance by reducing the noise in the recommendation process. Specifically, (1) for the noise in KG, we design a personalized gating mechanism, namely dual-channel balancing mechanism, to block the propagation of redundant information in KG. (2) For the noise in user-item interaction data, we integrate personalized and knowledge-aware signals to capture user preferences fully and use personalized knowledge-aware attention to denoise user-item interaction data. Compared with existing KG-based methods, we aim to propose a knowledge-aware recommendation method from a new perspective of denoising. We perform performance analysis on three real-world datasets, and experiment results demonstrate that KDGNN achieves strongly competitive performance compared with several compelling state-of-the-art baselines.

Dynamics analysis of the two-layer complex propagation network with individual heterogeneous decreased behavior

Due to the differences in society stratum, personal profession, and social acceptability, information propagation can be impacted by the contact capabilities of individuals. Importantly, we found that with the changes in individual psychology, their response to a phenomenon will gradually weaken. This phenomenon is called heterogeneous decreased behavior and applied in the fields of economics, sociology, and ecology. In the social network, people show a gradually decreasing degree of interest for information, named individual heterogeneous decreased behavior (IHDB). We structure a two-layer network model to describe individual behavioral contact and propose a threshold function to represent IHDB. Meanwhile, we use partition theory to explain the information propagation mechanism. Through experiments, it is demonstrated that there is a continuous information outbreak in the ultimate adoption size when individuals exhibit a positive IHDB. However, when individuals exhibit a passive IHDB, there is a discontinuous information outbreak in the ultimate adoption size. Eventually, our experiments show that the theoretical analysis coincides with the results of the simulations.

Consensus reaching process for group decision-making based on trust network and ordinal consensus measure

Due to the urgent nature of emergency decision making, it is necessary to reach the consensus requirement quickly. Ordinal consensus measure explores the relation between the rankings and helps to intuitively know which alternative needs to be adjusted to accelerate the improvement of consensus. Moreover, decision makers (DMs) in the decision making problem are often connected through trust relationships which affect the DMs’ judgments in the process of DMs’ interaction. Therefore, this paper explores trust network-based group decision-making in which the consensus level is estimated by an ordinal consensus measure. We first focus on the supplementation of an incomplete trust network. One of the most common methods is to design the trust propagation operator, whereas the intensity of information propagation may be different in various scenarios. Therefore, considering the different numerical scale of the linguistic term set, a trust propagation operator with different intensity of trust propagation is designed to obtain the indirect trust relationship. In the process of supplementing the incomplete trust network, the contribution of DMs to propagating information is concerned, which can be described by the betweenness centrality, and the importance weights of DMs are determined by combining the betweenness centrality and trust in-degree. In the consensus reaching process, we first propose an improved ordinal consensus measure, which takes into account the consistency of orders of the same alternative in different rankings as well as the importance of positions of alternatives. Then, we design the identification rule and the feedback mechanism for those with low consensus levels. The identification rule is used to select the DMs which first few alternatives in the ranking are different with those in the ranking of group. And in the feedback mechanism, the referenced preference relation (FPR) obtained by the trust network is provided for the identified DMs. Afterwards, combining the referenced FPR, an optimization model is designed to give the adjustment opinion. Finally, a numerical example elaborates on the feasibility of the trust propagation operator and consensus model. The comparative analysis demonstrates the rationality and effectiveness of the proposed model.

Daily runoff prediction based on the adaptive fourier decomposition method and multiscale temporal convolutional network.

The non-linearity and non-stationarity of runoff series pose significant challenges to runoff forecasting, and conventional single forecasting models struggle to accurately capture the internal dynamics of the series. To address this issue, we propose a runoff prediction model named AFDM-MTCN, which combines the adaptive Fourier decomposition method (AFDM) and multiscale temporal convolutional network (MTCN). AFDM-MTCN consists of two stages: adaptive decomposition and multi-scale feature extraction. In the adaptive decomposition stage, the improved Fourier decomposition method (IFDM) is optimized using the Sparrow Search Algorithm to enhance its ability to extract temporal patterns. In the multi-scale feature extraction stage, improvements are made to the temporal convolutional network (TCN) through the use of multi-scale convolution kernels, skip connections, and depth-wise separable convolution, to capture information from multiple angles, enhance information propagation, and reduce training parameters. The model was applied to two hydrological stations in the Weihe River Basin and compared with state-of-the-art methods to assess its accuracy and feasibility. The results demonstrate that AFDM-MTCN exhibits satisfactory performance in runoff prediction. Furthermore, compared to other decomposition techniques, AFDM demonstrates stronger capability in extracting patterns from non-stationary runoff data.

Deep stable neural networks: Large-width asymptotics and convergence rates

In modern deep learning, there is a recent and growing literature on the interplay between large-width asymptotics for deep Gaussian neural networks (NNs), i.e. deep NNs with Gaussian-distributed weights, and classes of Gaussian stochastic processes (SPs). Such an interplay has proved to be critical in several contexts of practical interest, e.g. Bayesian inference under Gaussian SP priors, kernel regression for infinite-wide deep NNs trained via gradient descent, and information propagation within infinite-wide NNs. Motivated by empirical analysis, showing the potential of replacing Gaussian distributions with Stable distributions for the NN's weights, in this paper we investigate large-width asymptotics for (fully connected) feed-forward deep Stable NNs, i.e. deep NNs with Stable-distributed weights. First, we show that as the width goes to infinity jointly over the NN's layers, a suitable rescaled deep Stable NN converges weakly to a Stable SP whose distribution is characterized recursively through the NN's layers. Because of the non-triangular NN's structure, this is a non-standard asymptotic problem, to which we propose a novel and self-contained inductive approach, which may be of independent interest. Then, we establish sup-norm convergence rates of a deep Stable NN to a Stable SP, quantifying the critical difference between the settings of ``joint growth and ``sequential growth of the width over the NN's layers. Our work extends recent results on infinite-wide limits for deep Gaussian NNs to the more general deep Stable NNs, providing the first result on convergence rates for infinite-wide deep NNs.

Influence Maximization in Complex Networks by Using Evolutionary Deep Reinforcement Learning

Influence maximization (IM) in complex networks tries to activate a small subset of seed nodes that could maximize the propagation of influence. The studies on IM have attracted much attention due to their wide applications such as item recommendation, viral marketing, information propagation and disease immunization. Existing works mainly model the IM problem as a discrete optimization problem, and use either approximate or meta-heuristic algorithms to address this problem. However, these works are hard to find a good tradeoff between effectiveness and efficiency due to the NP-hard and large-scale network properties of the IM problem. In this article, we propose an evolutionary deep reinforcement learning algorithm (called EDRL-IM) for IM in complex networks. First, EDRL-IM models the IM problem as a continuous weight parameter optimization of deep Q network (DQN). Then, it combines an evolutionary algorithm (EA) and a deep reinforcement learning algorithm (DRL) to evolve the DQN. The EA simultaneously evolves a population of individuals, and each of which represents a possible DQN and returns a solution to the IM problem through a dynamic markov node selection strategy, while the DRL integrates all information and network-specific knowledge of DQNs to accelerate their evolution. Systematic experiments on both benchmark and real-world networks show the superiority of EDRL-IM over the state-of-the-art IM methods in finding seed nodes.

FCD-AttResU-Net: An improved forest change detection in Sentinel-2 satellite images using attention residual U-Net

Forest Change Detection (FCD) is a critical component of natural resource monitoring and conservation strategies, enabling informed decision-making. Various methods utilizing the power of artificial intelligence (AI) have been developed for detecting and categorizing changes in forest cover using remote sensing (RS) data. One prominent AI-powered approach is the U-Net, a deep learning (DL) architecture famous for its segmentation proficiency. However, the standard U-Net architecture fails to effectively capture intricate spatial dependencies and long-range contextual information present in remote sensing imagery. To address this research gap, we introduce an attention-residual-based novel DL model which leverages the U-Net architecture and Sentinel-2 satellite images to map alterations in forest vegetation cover in the tropical region. Our novel model enhances the U-Net architecture by seamlessly integrating the strengths of the U-Net, harnessing attention mechanisms strategically to amplify crucial features, and leveraging cutting-edge residual connections to facilitate the smooth flow of information and gradient propagation. These meticulous design choices enabled the precise feature extraction, resulting in improved computational performance of the proposed method compared to the Standard U-Net, Deeplabv3+, Deep Res-U-Net, and Attention U-Net. The classification results demonstrate the enhanced efficiency of our model, achieving a Mean Intersection over Union (MIoU) of 0.9330 on our test dataset. This performance surpasses the Attention U-Net (0.9146), Standard U-Net (0.9029), Deeplabv3+ (0.9247), and Deep Res-U-Net (0.9282). The comparative analysis of ground truth reproductions unveiled the superior detection capabilities of our model in accurately identifying forest and non-forest polygons, surpassing both the standard U-Net, and the U-Net augmented with attention mechanism, along with other state-of-the-art techniques, thereby highlighting its enhanced efficacy. The model’s broad applicability can support forest managers and ecologists in rapidly evaluating the long-term ramifications of infrastructure initiatives, such as roads, on tropical forests, including those in Brunei.

Influential Node Detection and Ranking With Fusion of Heterogeneous Social Media Information

Identification of influential nodes has emerged as one of the major challenges, especially after their use in the rapid propagation of information, epidemics, and so on in social media. Most of the previous works in this field deal with the homogeneous interactions that are not pertinent in the determination of the accurate context of the nodes due to their noisy and sparse nature. Hence, heterogeneous interactions need to be explored for the identification of influential nodes in the network. To consider heterogeneous interactions available within the network, a multilayer network (ML) has been designed in this work. Each layer of the network represents a particular type of interaction, e.g., upload, comment, retweet, reply, and mention. A heterogeneous degree ranking (HDR)-based influential nodes’ detection and ranking are proposed for the designed ML. Furthermore, multiple-criteria decision-making (MCDM) methods, such as the analytic hierarchy process (AHP), the technique for order of preference by similarity to ideal solution (TOPSIS), fuzzy AHP, fuzzy TOPSIS, and the analytic network process (ANP), are explored for the proposed ML for identification and ranking of the influential nodes. The susceptible–infected–recovered (SIR) model is used to evaluate the proposed work. In addition to this, statistical analysis is performed using the Pearson correlation, Kendall’s correlation, Spearman’s correlation, and the Friedman test on the ranks generated by different methods, which shows that the results generated by different proposed methods are consistent. Furthermore, the performance of the proposed method is compared with state-of-the-art approaches.

Neighbor Library-Aware Graph Neural Network for Third Party Library Recommendation

Modern software development has moved toward agile growth and rapid delivery, where developers must meet the changing needs of users instantaneously. In such a situation, plug-and-play Third-Party Libraries (TPLs) introduce a considerable amount of convenience to developers. However, selecting the exact candidate that meets the project requirements from the countless TPLs is challenging for developers. Previous works have considered setting up a personalized recommender system to suggest TPLs for developers. Unfortunately, these approaches rarely consider the complex relationships between applications and TPLs, and are unsatisfactory in accuracy, training speed, and convergence speed. In this paper, we propose a new end-to-end recommendation model called Neighbor Library-Aware Graph Neural Network (NLA-GNN). Unlike previous works, we only initialize one type of node embedding, and construct and update all types of node representations using Graph Neural Networks (GNN). We use a simplified graph convolution operation to alternate the information propagation process to increase the training efficiency and eliminate the heterogeneity of the app-library bipartite graph, thus efficiently modeling the complex high-order relationships between the app and the library. Extensive experiments on large-scale real-world datasets demonstrate that NLA-GNN achieves consistent and remarkable improvements over state-of-the-art baselines for TPL recommendation tasks.

AN EVOLUTIONARY MODEL OF SOCIAL NETWORK STRUCTURE DRIVEN BY INFORMATION INTERACTION

The interaction of information and the evolution of network structure are inseparable. In order to construct social network evolution and information propagation models that better fit real-world scenarios, this paper proposes a social network structure evolution model driven by changes in the strength of relationships between individuals through their information interactions with each other. During the evolution process of the network, information interaction between individuals is also influenced by the network structure. Therefore, we improve traditional propagation models and construct an information propagation model with dynamic propagation rates. The proposed model is used to simulate both the spread of information and the evolution of network structures in real social networks. Finally, simulation results are compared to real-world data, demonstrating the effectiveness and rationality of the proposed model.

Illusion of diffusion: A percolation perspective on social structure and segregation in Twitter networks

Some diffusion trajectories produce cohesion while others generate highly fragmented social networks. Nevertheless, extant diffusion-based theoretical frameworks often assume continuous, gradual patterns of information flow and cascades and therefore cannot explain such disparity. Drawing on the percolation theory, which asserts that information propagation occurs abruptly and radically, this study conducts an empirical investigation into such structural integration or percolation efficiency in online social networks. In addition, we examine the roles of different characteristics of nodes and links, thereby identifying the detailed mechanism. The results demonstrate that node and link types significantly influence percolation efficiency. Regarding node types, influentials and broadcasters contribute to enhancing percolation efficiency, with the latter exhibiting a more substantial impact than the former. Hidden influentials exhibit a negligible effect on percolation efficiency, while common users negatively impact the efficiency of percolation. With respect to link types, incoming, outgoing, and bridge links serve to augment percolation efficiency. Overlap and adding links do not influence the efficiency of percolation, whereas new channel links impedes effective percolation. Based on a new theoretical perspective, this study extends research on information diffusion and word-of-mouth and provides stylized insights into effective social media targeting and advertising.

MASA-SegNet: A Semantic Segmentation Network for PolSAR Images

Semantic segmentation of Polarimetric SAR (PolSAR) images is an important research topic in remote sensing. Many deep neural network-based semantic segmentation methods have been applied to PolSAR image segmentation tasks. However, a lack of effective means to deal with the similarity of object features and speckle noise in PolSAR images exists. Thisstudy aims to improve the discriminative capability of neural networks for various intensities of backscattering coefficients while reducing the effects of noise in PolSAR semantic segmentation tasks. Firstly, we propose pre-processing methods for PolSAR image data, which consist of the fusion of multi-source data and false color mapping. Then, we propose a Multi-axis Sequence Attention Segmentation Network (MASA-SegNet) for semantic segmentation of PolSAR data, which is an encoder–decoder framework. Specifically, within the encoder, a feature extractor is designed and implemented by stacking Multi-axis Sequence Attention blocks to efficiently extract PolSAR features at multiple scales while mitigating inter-class similarities and intra-class differences from speckle noise. Moreover, the process of serialized residual connection design enables the propagation of spatial information throughout the network, thereby improving the overall spatial awareness of MASA-SegNet. Within the decoder, it is used to accomplish the semantic segmentation task. The superiority of this algorithm for semantic segmentation will be explored through feature visualization. The experiments show that our proposed spatial sequence attention mechanism can effectively extract features and reduce noise interference and is thus able to obtain the best results on two large-scale public datasets (the AIR-POlSAR-Seg and FUSAR-Map datasets).

Schooling of light reflecting fish.

One of the hallmarks of the collective movement of large schools of pelagic fish are waves of shimmering flashes that propagate across the school, usually following an attack by a predator. Such flashes arise when sunlight is reflected off the specular (mirror-like) skin that characterizes many pelagic fishes, where it is otherwise thought to offer a means for camouflage in open waters. While it has been suggested that these 'shimmering waves' are a visual manifestation of the synchronized escape response of the fish, the phenomenon has been regarded only as an artifact of esthetic curiosity. In this study we apply agent-based simulations and deep learning techniques to show that, in fact, shimmering waves contain information on the behavioral dynamics of the school. Our analyses are based on a model that combines basic rules of collective motion and the propagation of light beams in the ocean, as they hit and reflect off the moving fish. We use the resulting reflection patterns to infer the essential dynamics and inter-individual interactions which are necessary to generate shimmering waves. Moreover, we show that light flashes observed by the school members themselves may extend the range at which information can be communicated across the school. Assuming that fish pay heed to this information, for example by entering an apprehensive state of reduced response-time, our analysis suggests that it can speed up the propagation of information across the school. Further still, we use an artificial neural network to show that light flashes are, on their own, indicative of the state and dynamics of the school, and are sufficient to infer the direction of attack and the shape of the school with high accuracy.