PyTorchKR Papers

최신 논문

총 856편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2409.09887

Leiden-Fusion Partitioning Method for Effective Distributed Training of Graph Embeddings

In the area of large-scale training of graph embeddings, effective training frameworks and partitioning methods are critical for handling large networks. However, they face two major challenges: 1) existing synchronized distributed frameworks require continuous communication to access information from other machines, and 2) the inability of current partitioning methods to ensure that subgraphs remain connected components without isolated nodes, which is essential for effective training of GNNs since training relies on information aggregation from neighboring nodes. To address these issues, we introduce a novel partitioning method, named Leiden-Fusion, designed for large-scale training of graphs with minimal communication. Our method extends the Leiden community detection algorithm with a greedy algorithm that merges the smallest communities with highly connected neighboring communities. Our method guarantees that, for an initially connected graph, each partition is a densely connected subgraph with no isolated nodes. After obtaining the partitions, we train a GNN for each partition independently, and finally integrate all embeddings for node classification tasks, which significantly reduces the need for network communication and enhances the efficiency of distributed graph training. We demonstrate the effectiveness of our method through extensive evaluations on several benchmark datasets, achieving high efficiency while preserving the quality of the graph embeddings for node classification tasks.

arXiv:2409.10263

MDL-Pool: Adaptive Multilevel Graph Pooling Based on Minimum Description Length

Graph pooling compresses graphs and summarises their topological properties and features in a vectorial representation. It is an essential part of deep graph representation learning and is indispensable in graph-level tasks like classification or regression. Current approaches pool hierarchical structures in graphs by iteratively applying shallow pooling operators up to a fixed depth. However, they disregard the interdependencies between structures at different hierarchical levels and do not adapt to datasets that contain graphs with different sizes that may require pooling with various depths. To address these issues, we propose MDL-Pool, a pooling operator based on the minimum description length (MDL) principle, whose loss formulation explicitly models the interdependencies between different hierarchical levels and facilitates a direct comparison between multiple pooling alternatives with different depths. MDP-Pool builds on the map equation, an information-theoretic objective function for community detection, which naturally implements Occam's razor and balances between model complexity and goodness-of-fit via the MDL. We demonstrate MDL-Pool's competitive performance in an empirical evaluation against various baselines across standard graph classification datasets.

arXiv:2409.08004

Learning Communities from Equilibria of Nonlinear Opinion Dynamics

This paper studies community detection for a nonlinear opinion dynamics model from its equilibria. It is assumed that the underlying network is generated from a stochastic block model with two communities, where agents are assigned with community labels and edges are added independently based on these labels. Agents update their opinions following a nonlinear rule that incorporates saturation effects on interactions. It is shown that clustering based on a single equilibrium can detect most community labels (i.e., achieving almost exact recovery), if the two communities differ in size and link probabilities. When the two communities are identical in size and link probabilities, and the inter-community connections are denser than intra-community ones, the algorithm can achieve almost exact recovery under negative influence weights but fails under positive influence weights. Utilizing fixed point equations and spectral methods, we also propose a detection algorithm based on multiple equilibria, which can detect communities with positive influence weights. Numerical experiments demonstrate the performance of the proposed algorithms.

arXiv:2409.07725

GRE^2-MDCL: Graph Representation Embedding Enhanced via Multidimensional Contrastive Learning

Graph representation learning has emerged as a powerful tool for preserving graph topology when mapping nodes to vector representations, enabling various downstream tasks such as node classification and community detection. However, most current graph neural network models face the challenge of requiring extensive labeled data, which limits their practical applicability in real-world scenarios where labeled data is scarce. To address this challenge, researchers have explored Graph Contrastive Learning (GCL), which leverages enhanced graph data and contrastive learning techniques. While promising, existing GCL methods often struggle with effectively capturing both local and global graph structures, and balancing the trade-off between nodelevel and graph-level representations. In this work, we propose Graph Representation Embedding Enhanced via Multidimensional Contrastive Learning (GRE2-MDCL). Our model introduces a novel triple network architecture with a multi-head attention GNN as the core. GRE2-MDCL first globally and locally augments the input graph using SVD and LAGNN techniques. It then constructs a multidimensional contrastive loss, incorporating cross-network, cross-view, and neighbor contrast, to optimize the model. Extensive experiments on benchmark datasets Cora, Citeseer, and PubMed demonstrate that GRE2-MDCL achieves state-of-the-art performance, with average accuracies of 82.5%, 72.5%, and 81.6% respectively. Visualizations further show tighter intra-cluster aggregation and clearer inter-cluster boundaries, highlighting the effectiveness of our framework in improving upon baseline GCL models.

arXiv:2409.03916

A Survey on Signed Graph Embedding: Methods and Applications

A signed graph (SG) is a graph where edges carry sign information attached to it. The sign of a network can be positive, negative, or neutral. A signed network is ubiquitous in a real-world network like social networks, citation networks, and various technical networks. There are many network embedding models have been proposed and developed for signed networks for both homogeneous and heterogeneous types. SG embedding learns low-dimensional vector representations for nodes of a network, which helps to do many network analysis tasks such as link prediction, node classification, and community detection. In this survey, we perform a comprehensive study of SG embedding methods and applications. We introduce here the basic theories and methods of SGs and survey the current state of the art of signed graph embedding methods. In addition, we explore the applications of different types of SG embedding methods in real-world scenarios. As an application, we have explored the citation network to analyze authorship networks. We also provide source code and datasets to give future direction. Lastly, we explore the challenges of SG embedding and forecast various future research directions in this field.

arXiv:2409.00824

Accelerating Hybrid Agent-Based Models and Fuzzy Cognitive Maps: How to Combine Agents who Think Alike?

While Agent-Based Models can create detailed artificial societies based on individual differences and local context, they can be computationally intensive. Modelers may offset these costs through a parsimonious use of the model, for example by using smaller population sizes (which limits analyses in sub-populations), running fewer what-if scenarios, or accepting more uncertainty by performing fewer simulations. Alternatively, researchers may accelerate simulations via hardware solutions (e.g., GPU parallelism) or approximation approaches that operate a tradeoff between accuracy and compute time. In this paper, we present an approximation that combines agents who `think alike', thus reducing the population size and the compute time. Our innovation relies on representing agent behaviors as networks of rules (Fuzzy Cognitive Maps) and empirically evaluating different measures of distance between these networks. Then, we form groups of think-alike agents via community detection and simplify them to a representative agent. Case studies show that our simplifications remain accuracy.

arXiv:2408.16877공식 코드 · Papers with Code

Longitudinal Modularity, a Modularity for Link Streams

Temporal networks are commonly used to model real-life phenomena. When these phenomena represent interactions and are captured at a fine-grained temporal resolution, they are modeled as link streams. Community detection is an essential network analysis task. Although many methods exist for static networks, and some methods have been developed for temporal networks represented as sequences of snapshots, few works can handle link streams. This article introduces the first adaptation of the well-known Modularity quality function to link streams. Unlike existing methods, it is independent of the time scale of analysis. After introducing the quality function, and its relation to existing static and dynamic definitions of Modularity, we show experimentally its relevance for dynamic community evaluation.

arXiv:2408.11739

Network-based diversification of stock and cryptocurrency portfolios

Maintaining a balance between returns and volatility is a common strategy for portfolio diversification, whether investing in traditional equities or digital assets like cryptocurrencies. One approach for diversification is the application of community detection or clustering, using a network representing the relationships between assets. We examine two network representations, one based on a standard distance matrix based on correlation, and another based on mutual information. The Louvain and Affinity propagation algorithms were employed for finding the network communities (clusters) based on annual data. Furthermore, we examine building assets' co-occurrence networks, where communities are detected for each month throughout a whole year and then the links represent how often assets belong to the same community. Portfolios are then constructed by selecting several assets from each community based on local properties (degree centrality), global properties (closeness centrality), or explained variance (Principal component analysis), with three value ranges (max, med, min), calculated on a maximal spanning tree or a fully connected community sub-graph. We explored these various strategies on data from the S\&P 500 and the Top 203 cryptocurrencies with a market cap above 2M USD in the period from Jan 2019 to Sep 2022. Moreover, we study into more details the periods of the beginning of the COVID-19 outbreak and the start of the war in Ukraine. The results confirm some of the previous findings already known for traditional stock markets and provide some further insights, while they reveal an opposing trend in the crypto-assets market.

arXiv:2408.09072

Enhancing Community Detection in Networks: A Comparative Analysis of Local Metrics and Hierarchical Algorithms

The analysis and detection of communities in network structures are becoming increasingly relevant for understanding social behavior. One of the principal challenges in this field is the complexity of existing algorithms. The Girvan-Newman algorithm, which uses the betweenness metric as a measure of node similarity, is one of the most representative algorithms in this area. This study employs the same method to evaluate the relevance of using local similarity metrics for community detection. A series of local metrics were tested on a set of networks constructed using the Girvan-Newman basic algorithm. The efficacy of these metrics was evaluated by applying the base algorithm to several real networks with varying community sizes, using modularity and NMI. The results indicate that approaches based on local similarity metrics have significant potential for community detection.

arXiv:2407.08069

Herding Unmasked: Insights into Cryptocurrencies, Stocks and US ETFs

Herding behavior has become a familiar phenomenon to investors, with potential dangers of both undervaluing and overvaluing assets, while also threatening market stability. This study contributes to the literature on herding behavior by using a recent dataset, covering the most impactful events of recent years. To our knowledge, this is the first study examining herding behavior across three different types of investment vehicle and also the first study observing herding at a community (subset) level. Specifically, we first explore this phenomenon in each separate type of investment vehicle, namely stocks, US ETFs and cryptocurrencies, using the Cross-Sectional Absolute Deviation model. We find mostly similar herding patterns for stocks and US ETFs. Subsequently, the same experiment is implemented on a combination of all three investment vehicles. For a deeper investigation, we adopt graph-based techniques including the Minimum Spanning Tree and Louvain community detection to partition the combination into smaller subsets to detect herding behavior for each subset. We find that herding behavior exists at all times across all types of investment vehicle at a subset level, although perhaps not at the superset level, and that this herding behavior tends to stem from specific events that solely impact that subset of assets. Lastly, we explore herding by examining the financial contagion effects between these types of investment vehicle. Results show that US ETFs not only have a tendency to propagate similar trading behaviors in stocks and especially cryptocurrencies but also show self-reinforcing herding behavior, acting as drivers of their own trends.

arXiv:2407.02679

Enhancing Scalability of Optimal Kron-based Reduction of Networks (Opti-KRON) via Decomposition with Community Detection

Electrical networks contain thousands of interconnected nodes and edges, which leads to computational challenges in some power system studies. To address these challenges, we contend that network reductions can serve as a framework to enable scalable computing in power systems. By building upon a prior AC "Opti-KRON" formulation, this paper presents a DC power flow formulation for finding network reductions that are optimal within the context of large transmission analysis. Opti-KRON previously formulated optimal Kron-based network reductions as a mixed integer linear program (MILP), where the number of binary variables scaled with the number of nodes. To improve the scalability of the Opti-KRON approach, we augment the MILP formulation with a community detection (CD) technique that segments a large network into smaller, disjoint, but contiguous sub-graphs (i.e., communities). For each sub-graph, we then (in parallel) apply MILP-based along with a new cutting plane constraint, thus, enhancing scalability. Ultimately, the new DC-based Opti-KRON method can achieve a 80-95\% reduction of networks (in terms of nodes) while statistically outperforming other CD- and Kron-based methods. We present simulation results for the IEEE RTS-96 and the 2383-bus Polish networks.

arXiv:2406.17135공식 코드 · Papers with Code

Testing network clustering algorithms with Natural Language Processing

The advent of online social networks has led to the development of an abundant literature on the study of online social groups and their relationship to individuals' personalities as revealed by their textual productions. Social structures are inferred from a wide range of social interactions. Those interactions form complex -- sometimes multi-layered -- networks, on which community detection algorithms are applied to extract higher order structures. The choice of the community detection algorithm is however hardily questioned in relation with the cultural production of the individual they classify. In this work, we assume the entangled nature of social networks and their cultural production to propose a definition of cultural based online social groups as sets of individuals whose online production can be categorized as social group-related. We take advantage of this apparently self-referential description of online social groups with a hybrid methodology that combines a community detection algorithm and a natural language processing classification algorithm. A key result of this analysis is the possibility to score community detection algorithms using their agreement with the natural language processing classification. A second result is that we can assign the opinion of a random user at >85% accuracy.

arXiv:2406.17556공식 코드 · Papers with Code

Modularity Based Community Detection in Hypergraphs

In this paper, we propose a scalable community detection algorithm using hypergraph modularity function, h-Louvain. It is an adaptation of the classical Louvain algorithm in the context of hypergraphs. We observe that a direct application of the Louvain algorithm to optimize the hypergraph modularity function often fails to find meaningful communities. We propose a solution to this issue by adjusting the initial stage of the algorithm via carefully and dynamically tuned linear combination of the graph modularity function of the corresponding two-section graph and the desired hypergraph modularity function. The process is guided by Bayesian optimization of the hyper-parameters of the proposed procedure. Various experiments on synthetic as well as real-world networks are performed showing that this process yields improved results in various regimes.

arXiv:2406.15155

Introducing the modularity graph: an application to brain functional networks

In signal processing, exploring complex systems through network representations has become an area of growing interest. This study introduces the modularity graph, a new graph-based feature, to highlight the relationship across the graph communities. After showing an application to the random graph class known as Stochastic Block Model, we consider the brain functional connectivity network estimated from real EEG data. The modularity graph provides a quantitative framework for examining the interactions between neuron clusters within the brain's network. The modularity graph works alongside multiscale community detection algorithms, thereby enabling the identification of community structures at various scales. After introducing the modularity graph, we apply it to the brain functional connectivity network, estimated from publicly available EEG recordings of motor imagery experiments. Statistical analysis across multiple scales shows that the modularity graph differs for the distinct brain connectivity states associated with various motor imagery tasks. This work emphasizes the application of signal on graph processing techniques to understand brain behavior during specific cognitive tasks, leveraging the novel modularity graph to identify patterns of brain connectivity in different cognitive conditions. This approach sets the stage for further signal on graph analysis to devise brain network modularity, and to gain insights into the motor imagery mechanisms.

arXiv:2406.15142

Community Detection from Multiple Observations: from Product Graph Model to Brain Applications

This paper proposes a multilayer graph model for the community detection from multiple observations. This is a very frequent situation, when different estimators are applied to infer graph edges from signals at its nodes, or when different signal measurements are carried out. The multilayer network stacks the graph observations at the different layers, and it links replica nodes at adjacent layers. This configuration matches the Cartesian product between the ground truth graph and a path graph, where the number of nodes corresponds to the number of the observations. Stemming on the algebraic structure of the Laplacian of the Cartesian multilayer network, we infer a subset of the eigenvectors of the true graph and perform community detection. Experimental results on synthetic graphs prove the accuracy of the method, which outperforms state-of-the-art approaches in terms of ability of correctly detecting graph communities. Finally, we show the application of our method to discriminate different brain networks derived from real EEG data collected during motor imagery experiments. We conclude that our approach appears promising in identifying graph communities when multiple observations of the graph are available and it results promising for EEG-based motor imagery applications.

arXiv:2406.13889

Network-community analysis of cellular senescence

Most cellular phenotypes are genetically complex. Identifying the set of genes that are most closely associated with a specific cellular state is still an open question in many cases. Here we study the transcriptional profile of cellular senescence using a combination of network-based approaches, which include eigenvector centrality feature selection and community detection. We apply our method to cell-type-resolved RNA sequencing data obtained from injured muscle tissue in mice. The analysis identifies some genetic markers consistent with previous findings, and other previously unidentified ones, which are validated with previously published single-cell RNA sequencing data in a different type of tissue. The key identified genes, both those previously known and the newly identified ones, are transcriptional targets of factors known to be associated with established hallmarks of senescence, and can thus be interpreted as molecular correlates of such hallmarks. The method proposed here could be applied to any complex cellular phenotype even when only bulk RNA sequencing is available, provided the data is resolved by cell type.

arXiv:2406.14288공식 코드 · Papers with Code

Revisiting Modularity Maximization for Graph Clustering: A Contrastive Learning Perspective

Graph clustering, a fundamental and challenging task in graph mining, aims to classify nodes in a graph into several disjoint clusters. In recent years, graph contrastive learning (GCL) has emerged as a dominant line of research in graph clustering and advances the new state-of-the-art. However, GCL-based methods heavily rely on graph augmentations and contrastive schemes, which may potentially introduce challenges such as semantic drift and scalability issues. Another promising line of research involves the adoption of modularity maximization, a popular and effective measure for community detection, as the guiding principle for clustering tasks. Despite the recent progress, the underlying mechanism of modularity maximization is still not well understood. In this work, we dig into the hidden success of modularity maximization for graph clustering. Our analysis reveals the strong connections between modularity maximization and graph contrastive learning, where positive and negative examples are naturally defined by modularity. In light of our results, we propose a community-aware graph clustering framework, coined MAGI, which leverages modularity maximization as a contrastive pretext task to effectively uncover the underlying information of communities in graphs, while avoiding the problem of semantic drift. Extensive experiments on multiple graph datasets verify the effectiveness of MAGI in terms of scalability and clustering performance compared to state-of-the-art graph clustering methods. Notably, MAGI easily scales a sufficiently large graph with 100M nodes while outperforming strong baselines.

arXiv:2406.12297공식 코드 · Papers with Code

Faithful Density-Peaks Clustering via Matrix Computations on MPI Parallelization System

Density peaks clustering (DP) has the ability of detecting clusters of arbitrary shape and clustering non-Euclidean space data, but its quadratic complexity in both computing and storage makes it difficult to scale for big data. Various approaches have been proposed in this regard, including MapReduce based distribution computing, multi-core parallelism, presentation transformation (e.g., kd-tree, Z-value), granular computing, and so forth. However, most of these existing methods face two limitations. One is their target datasets are mostly constrained to be in Euclidian space, the other is they emphasize only on local neighbors while ignoring global data distribution due to restriction to cut-off kernel when computing density. To address the two issues, we present a faithful and parallel DP method that makes use of two types of vector-like distance matrices and an inverse leading-node-finding policy. The method is implemented on a message passing interface (MPI) system. Extensive experiments showed that our method is capable of clustering non-Euclidean data such as in community detection, while outperforming the state-of-the-art counterpart methods in accuracy when clustering large Euclidean data. Our code is publicly available at https://github.com/alanxuji/FaithPDP.

arXiv:2406.12412

Community Detection in Networks: A Rough Sets and Consensus Clustering Approach

The objective of this paper is to propose a framework, called Rough Clustering-based Consensus Community Detection (RC-CCD), to effectively address the challenge of identifying community structures in complex networks from a set of different community partitions. The method uses a consensus approach based on Rough Set Theory (RST) to manage uncertainty and improve the reliability of community detection. The RC-CCD framework is tested on synthetic benchmark networks generated by the Lancichinetti-Fortunato-Radicchi (LFR) method, which simulate varying network scales, node degrees, and community sizes. Key findings demonstrate that RC-CCD outperforms established algorithms like Louvain, Greedy, and LPA in terms of normalized mutual information, showing superior accuracy and adaptability, particularly in networks with higher complexity, both in terms of size and dispersion. These results have significant implications for enhancing community detection in fields such as social and biological network analysis.

arXiv:2406.07849

Bias-Corrected Joint Spectral Embedding for Multilayer Networks with Invariant Subspace: Entrywise Eigenvector Perturbation and Inference

In this paper, we propose to estimate the invariant subspace across heterogeneous multiple networks using a novel bias-corrected joint spectral embedding algorithm. The proposed algorithm recursively calibrates the diagonal bias of the sum of squared network adjacency matrices by leveraging the closed-form bias formula and iteratively updates the subspace estimator using the most recent estimated bias. Correspondingly, we establish a complete recipe for the entrywise subspace estimation theory for the proposed algorithm, including a sharp entrywise subspace perturbation bound and the entrywise eigenvector central limit theorem. Leveraging these results, we settle two multiple network inference problems: the exact community detection in multilayer stochastic block models and the hypothesis testing of the equality of membership profiles in multilayer mixed membership models. Our proof relies on delicate leave-one-out and leave-two-out analyses that are specifically tailored to block-wise symmetric random matrices and a martingale argument that is of fundamental interest for the entrywise eigenvector central limit theorem.