Matching Nodes Research Articles

PAGE: Prototype-Based Model-Level Explanations for Graph Neural Networks.

Aside from graph neural networks (GNNs) attracting significant attention as a powerful framework revolutionizing graph representation learning, there has been an increasing demand for explaining GNN models. Although various explanation methods for GNNs have been developed, most studies have focused on instance-level explanations, which produce explanations tailored to a given graph instance. In our study, we propose Prototype-bAsed GNN-Explainer ([Formula: see text]), a novel model-level GNN explanation method that explains what the underlying GNN model has learned for graph classification by discovering human-interpretable prototype graphs. Our method produces explanations for a given class, thus being capable of offering more concise and comprehensive explanations than those of instance-level explanations. First, [Formula: see text] selects embeddings of class-discriminative input graphs on the graph-level embedding space after clustering them. Then, [Formula: see text] discovers a common subgraph pattern by iteratively searching for high matching node tuples using node-level embeddings via a prototype scoring function, thereby yielding a prototype graph as our explanation. Using six graph classification datasets, we demonstrate that [Formula: see text] qualitatively and quantitatively outperforms the state-of-the-art model-level explanation method. We also carry out systematic experimental studies by demonstrating the relationship between [Formula: see text] and instance-level explanation methods, the robustness of [Formula: see text] to input data scarce environments, and the computational efficiency of the proposed prototype scoring function in [Formula: see text].

Pregnant People-Infant Linkage for Surveillance

Objectives and ApproachAs maternal mortality and morbidity continue to progress, better data is needed to comprehensively understand the problem, especially in underserved populations who are differentially impacted. Surveillance in safety-net populations is complicated by care fragmentation and interoperability challenges between primary care and hospital settings. . AllianceChicago (AC), a health center-controlled network, supports a common data infrastructure across a set of Federally Qualified Health Centers (FQHCs) in Chicago that is further linked to a distributed query infrastructure connecting Chicago’s major health systems known as CAPriCORN. Leveraging electronic health record (EHR) data collected by the AC and CAPriCORN consortium, the objective of this surveillance effort was to connect infant data with maternal data while protecting privacy. We developed a deterministic matching algorithm that utilized multiple approaches and technologies for Pregnant People-Infant (PPI) pair identification. The linkage process involved a five-step method in which each step is recorded within the linkage database and the confidence interval is calculated. ResultsPreliminary analyses are ongoing. Through each step of the linkage process, the confidence interval increased with the increase of matching nodes, highlighting its accumulative accuracy. The result demonstrates the feasibility of linking maternal and baby healthcare characteristics using a range of clinical and demographic variables captured in pseudonymized hospital and ambulatory data. Conclusions and ImplicationsOur approach improves the capacity of EHR data for surveillance while protecting patient privacy and preserving data confidentiality. Triangulating outcomes recorded in different settings can help improve data quality for research and surveillance.

SEMScene: Semantic-Consistency Enhanced Multi-Level Scene Graph Matching for Image-Text Retrieval

Image-text retrieval, a fundamental cross-modal task, performs similarity reasoning for images and texts. The primary challenge for image-text retrieval is cross-modal semantic heterogeneity, where the semantic features of visual and textual modalities are rich but distinct. Scene graph is an effective representation for images and texts as it explicitly models objects and their relations. Existing scene graph based methods have not fully taken the features regarding various granularities implicit in scene graph into consideration (e.g. triplets), the inadequate feature matching incurs the absence of non-trivial semantic information (e.g. inner relations among triplets). Therefore, we propose a S emantic-Consistency E nhanced M ulti-Level Scene Graph Matching (SEMScene) network, which exploits the semantic relevance between visual and textual scene graphs from fine-grained to coarse-grained. Firstly, under the scene graph representation, we perform feature matching including low-level node matching, mid-level semantic triplet matching, and high-level holistic scene graph matching. Secondly, to enhance the semantic-consistency for object-fused triplets carrying key correlation information, we propose a dual-step constraint mechanism in mid-level matching. Thirdly, to guide the model to learn the semantic-consistency of matched image-text pairs, we devise effective loss functions for each stage of the dual-step constraint. Comprehensive experiments on Flickr30K and MS-COCO datasets demonstrate that SEMScene achieves state-of-the-art performances with significant improvements.

Research on Computing Resource Measurement and Routing Methods in Software Defined Computing First Network.

Computing resource measurement and computing routing are essential technologies in the computing first network (CFN), serving as its foundational elements. This paper introduces a Software Defined Computing First Network (SD-CFN) architecture. Building upon this framework, a Dynamic-Static Integrated Computing Resource Measurement Mechanism (DCRMM) is proposed, incorporating methods such as the entropy weight method and K-Means clustering. The DCRMM algorithm outperforms the Maximum-closest Static Algorithm (MSA) and Maximum Closest Dynamic Algorithm (MDA) in terms of node stability, node utilization, and node matching accuracy. Additionally, a Reinforcement Learning and Software Defined Computing First Networking Routing (RSCR) algorithm is presented as a software-defined computing routing solution within the SD-CFN. RSCR introduces a knowledge plane responsible for computing routing calculations. It comprehensively considers factors such as link latency, available bandwidth, and packet loss rate. Simulation experiments conducted on the GÉANT topology demonstrate that RSCR outperforms the OSPF algorithm in terms of link latency, packet loss rate, and throughput. DCRMM and RSCR offer innovative solutions for computing resource measurement and computing routing in computing first networks.

On the Power of Gradual Network Alignment Using Dual-Perception Similarities.

Network alignment (NA) is the task of finding the correspondence of nodes between two networks based on the network structure and node attributes. Our study is motivated by the fact that, since most of existing NA methods have attempted to discover all node pairs at once, they do not harness information enriched through interim discovery of node correspondences to more accurately find the next correspondences during the node matching. To tackle this challenge, we propose Grad-Align, a new NA method that gradually discovers node pairs by making full use of node pairs exhibiting strong consistency, which are easy to be discovered in the early stage of gradual matching. Specifically, Grad-Align first generates node embeddings of the two networks based on graph neural networks along with our layer-wise reconstruction loss, a loss built upon capturing the first-order and higher-order neighborhood structures. Then, nodes are gradually aligned by computing dual-perception similarity measures including the multi-layer embedding similarity as well as the Tversky similarity, an asymmetric set similarity using the Tversky index applicable to networks with different scales. Additionally, we incorporate an edge augmentation module into Grad-Align to reinforce the structural consistency. Through comprehensive experiments using real-world and synthetic datasets, we empirically demonstrate that Grad-Align consistently outperforms state-of-the-art NA methods.

Subgraph-aware virtual node matching Graph Attention Network for entity alignment

With the vigorous development of knowledge graph, its construction has become a hot issue recently. However, most knowledge graphs are incomplete, and it is worthwhile to devote much effort to further fuse the knowledge of different entities. Hence, entity alignment has surprisingly emerged as a key step of knowledge fusion to tackle the above issue. The existing GNN-based models simply aggregate entity embeddings, or incorporate auxiliary information of entities (e.g., relationship, time, etc.), which frustratingly fail to consider neighborhood structure of entities and the cross-graph matching information of the nodes between two knowledge graphs. Thus, to overcome the two challenges, we propose a novel Subgraph-aware Virtual Node Matching Graph Attention neTwork for entity alignment, named SVNM-GAT. Specifically, to strikingly capture the cross-graph matching interaction information of entities, we propose a new virtual node matching mechanism, which can obtain the long-range dependencies between virtual nodes and holistic nodes by scale-dot attention. Additionally, a subgraph discriminator is presented in SVNM-GAT to identify the subgraph structure and obtain weighted adjacent matrix by calculating subgraph structure coefficients. Finally, we adopt experiments to verify the effectiveness of our model on the cross-linguistic and temporal datasets, and all the results indicate that SVNM-GAT considerably outperforms the state-of-the-art methods.

Tracking Age-Related Topological Changes in Individual Brain Morphological Networks Across the Human Lifespan.

Many studies have shown topological alterations associated with age in population-based brain morphological networks. However, it is not clear how individual brain morphological networks change with age across the lifespan. To characterize age-related topological changes in individual networks and investigate the relationships between individual- and group-based brain networks at the nodal, modular, and connectome levels. Retrospective analysis. One hundred seventy-nine healthy subjects (108 males and 71 females), aged 6-85 years with a median age of 32 years and an inter-quartile range (IQR) of 26 years. T1-weighted images using the magnetization-prepared rapid gradient echo (MPRAGE) sequences. Two nodal-level indicators (nodal similarity and node matching), five modular-level indicators (modularity, intra/inter-module similarity, adjusted mutual information [AMI], and module variation), and five connectome-level indicators (global efficiency, characteristic path length, clustering coefficient, local efficiency, and individual contribution) were calculated in brain morphological networks. Regression models for different indicators were built to examine their lifetime trajectory patterns. Single-sample t-test, Mantel's test, Pearson correlation coefficient. A P value <0.05 was considered statistically significant. Among 68 nodes, 34 nodes showed significant age-related patterns (all P < 0.05, FDR-corrected) in nodal similarity, including linear decline and quadratic trends. The lifespan trajectory of the connectome-level topological attributes of the individual networks presented U-shaped or inverse U-shaped trends with age. Between the individual- and group-based brain networks, the average nodal similarity was 0.67 and the average AMI of module partitions was 0.57. The lifespan trajectories of the nodal similarity mainly followed linear decreasing and nonlinear trends, whereas the modularity and the global topological attributes exhibited nonlinear patterns. There was a high degree of consistency at both nodal similarity and modular division between the individual and group networks. 1 TECHNICAL EFFICACY: Stage 1.

Reconversion of a building to a new functional program using graph matching

Building, demolishing, and rebuilding new structures, instead of renovating vacant buildings for adaptive reuse, generates a significant amount of waste. Quickly assessing if and how a building can house a new function can be time-consuming. We developed a method that automates this process using a graph-based strategy in which both the building and the functional program are represented by attributed graphs. Specifically, the new method performs a node matching strategy based on the properties of the nodes and the topological structure of the graphs by which each empty room is filled with an activity of the new function.

FedDNA: Federated learning using dynamic node alignment.

Federated Learning (FL), as a new computing framework, has received significant attentions recently due to its advantageous in preserving data privacy in training models with superb performance. During FL learning, distributed sites first learn respective parameters. A central site will consolidate learned parameters, using average or other approaches, and disseminate new weights across all sites to carryout next round of learning. The distributed parameter learning and consolidation repeat in an iterative fashion until the algorithm converges or terminates. Many FL methods exist to aggregate weights from distributed sites, but most approaches use a static node alignment approach, where nodes of distributed networks are statically assigned, in advance, to match nodes and aggregate their weights. In reality, neural networks, especially dense networks, have nontransparent roles with respect to individual nodes. Combined with random nature of the networks, static node matching often does not result in best matching between nodes across sites. In this paper, we propose, FedDNA, a dynamic node alignment federated learning algorithm. Our theme is to find best matching nodes between different sites, and then aggregate weights of matching nodes for federated learning. For each node in a neural network, we represent its weight values as a vector, and use a distance function to find most similar nodes, i.e., nodes with the smallest distance from other sides. Because finding best matching across all sites are computationally expensive, we further design a minimum spanning tree based approach to ensure that a node from each site will have matched peers from other sites, such that the total pairwise distances across all sites are minimized. Experiments and comparisons demonstrate that FedDNA outperforms commonly used baseline, such as FedAvg, for federated learning.

Offloading Strategy of Multi-Service and Multi-User Edge Computing in Internet of Vehicles

An edge computing offloading strategy was proposed with the goal of addressing the issue of low edge computing efficiency and service quality in the multi-service and multi-user intersections of networked vehicles. This strategy took into account all relevant factors, including the matching of users and service nodes, offloading ratio, bandwidth and computing power resource allocation, and system energy consumption. It is mainly divided into 2 tasks: (1) Service node selection: A fuzzy logic-based service node selection algorithm (SNFLC) is proposed. The linear equation for node performance value is determined through fuzzy reasoning by specifying three performance indexes as input. Gradient descent method is used to find the optimal value of the objective function, and the Lyapunov criterion coefficient is introduced to improve the stability of the algorithm. (2) Offload ratio and resource allocation are solved: The coupling between offload ratio and bandwidth resource allocation is confirmed by relaxing integer variables because the optimization goal problem is a NP problem, and the issue is divided into two sub-problems. At the same time, a low-complexity alternate iteration resource allocation algorithm (LC-IRA) is proposed to solve the bandwidth resource and computational power resource allocation. According to simulation results, the performance of genetic ant colony algorithm (G_ACA), non orthogonal multiple access technology (NOMA) and LC-IRA are improved by 26.5%, 31.37%, and 45.52%, respectively, compared with the random unloading allocation (RUA) and average distribution (AD).

A multi-level generative framework for community detection in attributed networks

AbstractCommunity detection in attributed networks is one of the most important tasks in complex network analysis. Many existing methods propose to integrate the network topology and node attribute from a generative aspect, which models an attributed network as a probabilistic generation process with the community distribution described by hidden variables. Though they can provide good interpretability to the community structure, it is difficult to infer community membership quickly due to their high computational complexity when inferring. Motivated by the multi-level strategy, in this study, we propose a multi-level generative framework to reduce the time cost of generative models for community detection in attributed networks. We first coarsen an attributed network into smaller ones by node matching. Then, we employ the existing generative model on the coarsest network without any modification for community detection, thus efficiently obtaining community memberships of nodes in this small coarsest network. Last, we project the assignments back to the original network through a local refinement mechanism to get communities. Extensive experiments on several real-world and artificial attributed networks show that our multi-level-based method is significantly faster than original generative models and is able to achieve better or more competitive results.

Unexpected Aspects of Expectancy in Music

A well agreed discourse in music perception research is that affective response can be generated by music when a tendency in the music is delayed or inhibited. There is a consensus that this tendency is statistically driven, derived from exposure to culturally situated musical idioms. By presenting a neural-network inspired spreading activation model (SAM) this paper argues that the nature of the tendency is worthy of further investigation. SAM organises the music stream perceived by the listener continuously into segments such that a match with an existing ‘mental representation’ (node) is found, which is then linked to the node for the previously segmented part of the music stream, with the link between these nodes strengthening and consolidating with exposure. The currently activated segment (the music being sounded) will prime the best matching (strongest linked) node available, generating expectancy. Expectancy is defined as the most strongly primed segment, and emerges dynamically through experience with environmental and musical contexts, rather than schematic or prototypical means. Expectancy is the specific exemplar instance that the activated (currently sounding) segment of music and contextual factors prime. This hypothesis of veridical dominance has implications for enduring aspects of music expectancy theory: (1) individual experiences matter in the formation of expectations; (2) expectations are a dynamic process, that change and are updated with experience; (3) context plays a critical role in expectancy; and (4) schema, prototypes and statistical accounts of expectation should be treated as convenient approximations of underlying cognitive processes.

Parallel Overlapping Community Detection Algorithm on GPU

Community detection is one of the most representative graph mining applications, which is often assembled as a concurrent graph partition application to explore the maximum modularity (or gained modularity) of each community. However, many branch divergence operations create significant obstacles to unleashing GPU&#x2019;s high throughput and memory bandwidth, which are needed in community detection applications to divide the vertices into different communities. In this paper, we present Lugger, a GPU-based overlapping community detection algorithm that reduces GPU&#x2019;s branch divergence via the customer-designed cache-aware parallel searching technique. In Lugger, we first design a cache-aware parallel searching policy using the B-Tree structure. Then, we set the B-Tree node matches with the GPU cache line to meet the coalesced memory access manner and avoid the branch divergence in warps. Moreover, we design a positive node splitting scheme to reduce the lock operation and idle threads when building the B-Tree structure. In addition, we implement a warp-centric thread assignment strategy to make sure the workloads across threads are balanced. We implement the proposed algorithm on NVIDIA GPU and evaluate the performance on eight large graphs (up to <inline-formula><tex-math notation="LaTeX">$3M$</tex-math></inline-formula> vertices and <inline-formula><tex-math notation="LaTeX">$117M$</tex-math></inline-formula> edges) with ground-truth communities. The experimental results show that Lugger can outperform the state-of-the-art works on scalability and detection quality.

Predicting Potential Drug-Disease Associations Based on Hypergraph Learning with Subgraph Matching.

The search for potential drug-disease associations (DDA) can speed up drug development cycles, reduce costly wasted resources, and accelerate disease treatment by repurposing existing drugs that can control further disease progression. As technologies such as deep learning continue to mature, many researchers tend to use emerging technologies to predict potential DDA. The performance of DDA prediction is still challenging and there is some space for improvement due to issues such as the small number of existing associations and possible noise in the data. To better predict DDA, we propose a computational approach based on hypergraph learning with subgraph matching (HGDDA). In particular, HGDDA first extracts feature subgraph information in the validated drug-disease association network and proposes a negative sampling strategy based on similarity network to reduce the data imbalance. Second, the hypergraph Unet module is used by extracting Finally, the potential DDA is predicted by designing a hypergraph combination module to convolution and pooling the two constructed hypergraphs separately, and calculating the difference information between the subgraphs using cosine similarity for node matching. The performance of HGDDA is verified under two standard datasets by 10-fold cross-validation (10-CV), and the results outperform existing drug-disease prediction methods. In addition, to validate the overall utility of the model, the top 10 drugs for the specific disease are predicted through the case study and validated using the CTD database.

A Trie Based Set Similarity Query Algorithm

Set similarity query is a primitive for many applications, such as data integration, data cleaning, and gene sequence alignment. Most of the existing algorithms are inverted index based, they usually filter unqualified sets one by one and do not have sufficient support for duplicated sets, thus leading to low efficiency. To solve this problem, this paper designs T-starTrie, an efficient trie based index for set similarity query, which can naturally group sets with the same prefix into one node, and can filter all sets corresponding to the node at a time, thereby significantly improving the candidates generation efficiency. In this paper, we find that the set similarity query problem can be transformed into matching nodes of the first-layer (FMNodes) detecting problem on T-starTrie. Therefore, an efficient FLMNode detection algorithm is designed. Based on this, an efficient set similarity query algorithm, TT-SSQ, is implemented by developing a variety of filtering techniques. Experimental results show that TT-SSQ can be up to 3.10x faster than existing algorithms.

SEEN: Seen: Sharpening Explanations for Graph Neural Networks Using Explanations From Neighborhoods

Explaining the foundations for predictions obtained from graph neural networks (GNNs) is critical for credible use of GNN models for real-world problems. Owing to the rapid growth of GNN applications, recent progress in explaining predictions from GNNs, such as sensitivity analysis, perturbation methods, and attribution methods, showed great opportunities and possibilities for explaining GNN predictions. In this study, we propose a method to improve the explanation quality of node classification tasks that can be applied in a post hoc manner through aggregation of auxiliary explanations from important neighboring nodes, named SEEN. Applying SEEN does not require modification of a graph and can be used with diverse explainability techniques due to its independent mechanism. Experiments on matching motifparticipating nodes from a given graph show great improvement in explanation accuracy of up to 12.71% and demonstrate the correlation between the auxiliary explanations and the enhanced explanation accuracy through leveraging their contributions. SEEN provides a simple but effective method to enhance the explanation quality of GNN model outputs, and this method is applicable in combination with most explainability techniques.

AEDNet: Adaptive Edge-Deleting Network For Subgraph Matching

Subgraph matching is to find all subgraphs in a data graph that are isomorphic to an existing query graph. Subgraph matching is an NP-hard problem, yet has found its applications in many areas. Many learning-based methods have been proposed for graph matching, whereas few have been designed for subgraph matching. The subgraph matching problem is generally more challenging, mainly due to the different sizes between the two graphs, resulting in considerable large space of solutions. Also the extra edges existing in the data graph connecting to the matched nodes may lead to two matched nodes of two graphs having different adjacency structures and often being identified as distinct objects. Due to the extra edges, the existing learning based methods often fail to generate sufficiently similar node-level embeddings for matched nodes. This study proposes a novel Adaptive Edge-Deleting Network (AEDNet) for subgraph matching. The proposed method is trained in an end-to-end fashion. In AEDNet, a novel sample-wise adaptive edge-deleting mechanism removes extra edges to ensure consistency of adjacency structure of matched nodes, while a unidirectional cross-propagation mechanism ensures consistency of features of matched nodes. We applied the proposed method on six datasets with graph sizes varying from 20 to 2300. Our evaluations on six open datasets demonstrate that the proposed AEDNet outperforms six state-of-the-arts and is much faster than the exact methods on large graphs.

Integrating Station-Area Development with Rail Transit Networks: Lessons from Japan Railway in Tokyo

Rail transit is a primary transport mode in Tokyo and has become the backbone of its urban development. Tokyo rail transit networks consist of the subway in the central area, private railways in the suburban area, and Japan Railway (JR) in both central and suburban areas. Prior research revealed that JR has better coordination of transit node functions and station-area development than subway and private railway stations, but how Tokyo's JR station-area development integrates with its network growth is less explored. This work investigates how JR station-area development grows in tandem with network expansion, by revisiting different historical phases. The results show that, in the beginning of the twentieth century, JR connected major urban nodes in both central and suburban Tokyo, and therefore became the skeleton of Tokyo's transit systems. During the post-earthquake and post-war period, central subway and suburban private railways gradually met JR on the Yamanote line, enabling the major interchange stations to develop into urban centers. In the 1960s, JR increased its services to alleviate passenger congestion and undertook grade separation projects to reduce conflicts with road networks. The 1987 JR privatization marked a turning point of massive station-area redevelopment. JR station areas have been transformed from underutilized industrial land into high-density commercial use, and major JR stations were comprehensively replanned for identity and place-making. After a centennial development, JR stations have become important interchange and express-service nodes on Tokyo transit networks with intensive land use and different urban functions. The Tokyo JR case can provide insights for metropolitan cities on how to integrate rail transit infrastructure with station-area development, including matching critical nodes in transit networks with urban functions and high-density land use and connecting these urban functions through express line services.

High-Accuracy Nodal Staging of Head and Neck Cancer With USPIO-Enhanced MRI: A New Reading Algorithm Based on Node-to-Node Matched Histopathology.

Ultrasmall superparamagnetic iron oxide (USPIO)-enhanced magnetic resonance imaging (MRI) is a potential diagnostic tool for lymph node assessment in patients with head and neck cancer. Validation by radiologic-pathologic correlation is essential before the method is evaluated in clinical studies. In this study, MRI signal intensity patterns of lymph nodes are correlated to their histopathology to develop a new USPIO-enhanced MRI reading algorithm that can be used for nodal assessment in head and neck cancer patients. Ten head and neck cancer patients underwent in vivo USPIO-enhanced MRI before neck dissection. An ex vivo MRI of the neck dissection specimen was performed for precise coregistration of in vivo MRI with histopathology. Normal clinical histopathological workup was extended with meticulous matching of all lymph nodes regarded as potentially metastatic based on their in vivo MRI signal intensity pattern. On the basis of histopathology of resected nodes, in vivo MRI signal characteristics were defined separating benign from malignant lymph nodes. Fifteen of 34 node-to-node correlated lymph nodes with remaining signal intensity on T2*-weighted MRI were histopathologically metastatic and 19 were benign. Radiological analysis revealed that metastatic lymph nodes showed equal or higher MRI signal intensity when compared with lipid tissue on T2*-weighted MGRE sequence (15/16 lymph nodes; 94%), whereas healthy lymph nodes showed lower (17/19 lymph nodes; 89%) or complete attenuation of signal intensity (273/279; 98%) when compared with lipid tissue on T2*-weighted MGRE. Histopathology of all resected specimens identified 392 lymph nodes. Six lymph nodes with (micro)metastases were missed with in vivo MRI. Whether these 6 lymph nodes were correlated to a nonmalignant lymph node on in vivo MRI or could not be detected at all is unclear. We developed a new reading algorithm to differentiate benign from malignant lymph nodes in head and neck cancer patients on the basis of their appearance on high-resolution T2*-weighted USPIO-enhanced MRI. Next steps involve validation of our reading algorithm to further improve the accuracy of neck lymph node staging with USPIO-enhanced MRI in prospective clinical studies with larger number of patients.

SeqPolar: Sequence Matching of Polarized LiDAR Map With HMM for Intelligent Vehicle Localization

3D LiDAR maps are playing a more and more important role in intelligent vehicle localization. In this paper, the proposed SeqPolar localization method consists of two parts. The first part is a novel node-based LiDAR map representation method, termed polarized LiDAR map (PLM), and the second is a map matching algorithm based on a second-order hidden Markov model (HMM2) for vehicle localization. Unlike existing 3D LiDAR map, which is usually constructed by accumulating 3D LiDAR clouds collected at different times, the proposed polarized LiDAR map is generated from a set of nodes. Each node consists of three elements: a polarized LiDAR image, scene features extracted from the polarized LiDAR image, and sensor pose. The polarized LiDAR image encodes 3D coordinates and reflectivities of the 3D LiDAR cloud using a multi-channel image format, a more concise, straightforward, and structured representation of the 3D LiDAR cloud. In the localization step, we propose an HMM2-based method to match a sequence of input polarized images within the map nodes and find the nearest node from the PLM. Afterward, the vehicle is readily localized from the matched map node with 3D registration. The proposed SeqPolar localization method has been validated with the actual field dataset and the public KITTI database. Experimental results demonstrate that the proposed HMM2-based matching method can achieve up to 98% accuracy to find the nearest node from PLM. Moreover, the SeqPolar localization method based on the pre-built PLM can achieve 30-centimeter localization accuracy in average on both test datasets.