Analysts who explore high-dimensional data usually want three answers at once: Which samples belong together, how close the resulting groups are, and who influences whom accordingly. Classical clustering provides only hard labels, hiding both inter-cluster affinities and correlation flow. We introduce Hamiltonian Cycle Clustering with Asymmetric Correlation HCC-AC, a framework that converts the clustering task into an interpretable map where structure and directionality are visible at a single glance. HCC-AC first learns soft memberships by optimizing a joint global–local loss, preserving manifold structure while turning each label into a probability. These probabilities drive a Hamiltonian-cycle embedding: cluster anchors are ordered by affinity and placed evenly on a circle; samples fall radially towards their most-likely anchor, so clusters, their similarities (arc lengths), and outliers emerge immediately. Directed arrows connect anchors, their lengths showing correlation strength, transforming the map into a legible narrative of influence. Experiments on five benchmark datasets demonstrate that HCC-AC improves the knowledge discovery in clustering, i.e., indexes the clustering results, flags outliers reliably, and uncovers correlation pathways.

Adapters, a plug-in neural network module with some tunable parameters, have emerged as a parameter-efficient transfer learning technique for adapting pre-trained models to downstream tasks, especially for natural language processing (NLP) and computer vision (CV) fields. Meanwhile, learning recommendation models directly from raw item modality features --- e.g., texts of NLP and images of CV --- can enable effective and transferable recommender systems (called TransRec). In view of this, a natural question arises:can adapter-based learning techniques achieve parameter-efficient TransRec with good performance?

3D single object tracking in LiDAR point clouds (LiDAR SOT) plays a crucial role in autonomous driving. Current approaches all follow the Siamese paradigm based on appearance matching. However, LiDAR point clouds are usually textureless and incomplete, which hinders effective appearance matching. Besides, previous methods greatly overlook the critical motion clues among targets. In this work, beyond 3D Siamese tracking, we introduce amotion-centric paradigmto handle LiDAR SOT from a new perspective. Following this paradigm, we propose a matching-free two-stage trackerM$^{2}$2-Track. At the 1st-stage,$M^{2}$-Track localizes the target within successive frames viamotion transformation. Then it refines the target box throughmotion-assisted shape completion at the 2nd-stage. Due to the motion-centric nature, our method shows its impressive generalizability with limited training labels and provides good differentiability for end-to-end cycle training. This inspires us to explore semi-supervised LiDAR SOT by incorporating a pseudo-label-based motion augmentation and a self-supervised loss term. Under the fully-supervised setting, extensive experiments confirm that$M^{2}$-Track significantly outperforms previous state-of-the-arts on three large-scale datasets while running at57FPS($\sim$∼3%,$\sim$∼11%and$\sim$∼22%precision gains on KITTI, NuScenes, and Waymo Open Dataset respectively). While under the semi-supervised setting, our method performs on par with or even surpasses its fully-supervised counterpart using fewer than half labels from KITTI. Further analysis verifies each component's effectiveness and shows the motion-centric paradigm's promising potential for auto-labeling and unsupervised domain adaptation.

Numerous physical objects in our daily lives are grouped or ranked according to a stereotyped presentation style. For example, in a library, books are typically grouped and ranked based on classification numbers. However, for better comparison, we often need to re-group or re-rank the books using additional attributes such as ratings, publishers, comments, publication years, keywords, prices, etc., or a combination of these factors. In this article, we propose a novel mobile DR/MR-based application framework named DRCmpVis to achieve in-context multi-attribute comparisons of physical objects with text labels or textual information. The physical objects are scanned in the real world using mobile cameras. All scanned objects are then segmented and labeled by a convolutional neural network and replaced (diminished) by their virtual avatars in a DR environment. We formulate three visual comparison strategies, including filtering, re-grouping, and re-ranking, which can be intuitively, flexibly, and seamlessly performed on their avatars. This approach avoids breaking the original layouts of the physical objects. The computation resources in virtual space can be fully utilized to support efficient object searching and multi-attribute visual comparisons. We demonstrate the usability, expressiveness, and efficiency of DRCmpVis through a user study, NASA TLX assessment, quantitative evaluation, and case studies involving different scenarios.

Dimension reduction (DR) is commonly utilized to capture the intrinsic structure and transform high-dimensional data into low-dimensional space while retaining meaningful properties of the original data. It is used in various applications, such as image recognition, single-cell sequencing analysis, and biomarker discovery. However, contemporary parametric-free and parametric DR techniques suffer from several significant shortcomings, such as the inability to preserve global and local features and the poor generalisation performance. On the other hand, regarding explainability, it is crucial to comprehend the embedding process, especially the contribution of each part to the embedding process, while understanding how each feature affects the embedding results that identify critical components and help diagnose the embedding process. To address these problems, we have developed a deep neural network method called DMT-EV, which provides not only excellent performance in structural maintainability but also explainability to the DR therein. DMT-EV starts with data augmentation and a manifold-based loss function to improve embedding performance. The explanation is based on saliency maps and aims to examine the trained DMT-EV parameters and contributions of components during the embedding process. The proposed techniques are integrated with a visual interface to help the user to adjust DMT-EV to achieve better DR performance and explainability. The interactive visual interface makes it easier to illustrate the data features, compare different DR techniques, and investigate DR. An in-depth experimental comparison shows that DMT-EV consistently outperforms the state-of-the-art methods in both performance measures and explainability.

Clustering aims to group unlabelled samples based on their similarities and is widespread in high-dimensional data analysis. However, most of the clustering methods merely generate pseudo labels and thus are unable to simultaneously present the similarities between different clusters and outliers. This paper proposes a new framework called High-dimensional Clustering onto Hamiltonian Cycle (HCHC) to solve the above problems. First, HCHC combines global structure with local structure in one objective function for deep clustering, improving the labels as relative probabilities, to mine the similarities between different clusters while keeping the local structure in each cluster. Then, the anchors of different clusters are sorted on the optimal Hamiltonian cycle generated by the cluster similarities and mapped on the circumference of a circle. Finally, a sample with a higher probability of a cluster will be mapped closer to the corresponding anchor. In this way, our framework allows us to appreciate three aspects visually and simultaneously - clusters (formed by samples with high probabilities), cluster similarities (represented as circular distances), and outliers (recognized as dots far away from all clusters). The theoretical analysis and experiments illustrate the superiority of HCHC.

Node-link diagrams are the most popular form for graph visualization. Yet, salient information of a node-link diagram cannot be fully depicted by solely presenting the visualization. We propose to augment node-link diagrams by creating textual descriptions for interested information. We conduct an expert review and a user interview to identify six requirements of generated interpretations, including three requirements for connection extraction and three requirements for visual expression. Our solution, GraphDescriptor, generates textual descriptions with two stages: feature extraction and description generation. The first one identifies and extracts features of node-link diagrams, like node connections, visual designs, and types of graph layouts. The second stage creates a group of hierarchical sentences based on a pre-defined schema. To the best of our knowledge, our approach is the first attempt to generate textual descriptions automatically. Three use cases and the in-lab user study confirm the superiority of our approach.

An exemplar is an entity that represents a desirable instance in a multi-attribute configuration space. It offers certain strengths in some of its attributes without unduly compromising the strengths in other attributes. Exemplars are frequently sought after in real life applications, such as systems engineering, investment banking, drug advisory, product marketing and many others. We study a specific method for the visualization of multi-attribute configuration spaces, the Data Context Map (DCM), for its capacity in enabling users to identify proper exemplars. The DCM produces a 2D embedding where users can view the data objects in the context of the data attributes. We ask whether certain graphical enhancements can aid users to gain a better understanding of the attribute-wise tradeoffs and so select better exemplar sets. We conducted several user studies for three different graphical designs, namely iso-contour, value-shaded topographic rendering and terrain topographic rendering, and compare these with a baseline DCM display. As a benchmark we use an exemplar set generated via Pareto optimization which has similar goals but unlike humans can operate in the native high-dimensional data space. Our study finds that the two topographic maps are statistically superior to both the iso-contour and the DCM baseline display.

3D single object tracking (3D SOT) in LiDAR point clouds plays a crucial role in autonomous driving. Current approaches all follow the Siamese paradigm based on appearance matching. However, LiDAR point clouds are usually textureless and incomplete, which hinders effective appearance matching. Besides, previous methods greatly overlook the critical motion clues among targets. In this work, beyond 3D Siamese tracking, we introduce a motion-centric paradigm to handle 3D SOT from a new perspective. Following this paradigm, we propose a matching-free two-stage tracker M2-Track. At the 1st-stage, M2-Track localizes the target within successive frames via motion transformation. Then it refines the target box through motion-assisted shape completion at the 2nd-stage. Extensive experiments confirm that M2-Track significantly outperforms previous state-of-the-arts on three large-scale datasets while running at 57FPS (~ 8%, ~ 17% and ~ 22% precision gains on KITTI, NuScenes, and Waymo Open Dataset respectively). Further analysis verifies each component's effectiveness and shows the motioncentric paradigm's promising potential when combined with appearance matching. Code will be made available at https://github.com/Ghostish/Open3DSOT.

By skeptics and undecided we refer to nodes in clustered social networks that cannot be assigned easily to any of the clusters. Such nodes are typically found either at the interface between clusters (the undecided) or at their boundaries (the skeptics). Identifying these nodes is relevant in marketing applications like voter targeting, because the persons represented by such nodes are often more likely to be affected in marketing campaigns than nodes deeply within clusters. So far this identification task is not as well studied as other network analysis tasks like clustering, identifying central nodes, and detecting motifs. We approach this task by deriving novel geometric features from the network structure that naturally lend themselves to an interactive visual approach for identifying interface and boundary nodes.

A wide variety of color schemes have been devised for mapping scalar data to color. We address the challenge of color-mapping multivariate data. While a number of methods can map low-dimensional data to color, for example, using bilinear or barycentric interpolation for two or three variables, these methods do not scale to higher data dimensions. Likewise, schemes that take a more artistic approach through color mixing and the like also face limits when it comes to the number of variables they can encode. Our approach does not have these limitations. It is data driven in that it determines a proper and consistent color map from first embedding the data samples into a circular interactive multivariate color mapping display (ICD) and then fusing this display with a convex (CIE HCL) color space. The variables (data attributes) are arranged in terms of their similarity and mapped to the ICD's boundary to control the embedding. Using this layout, the color of a multivariate data sample is then obtained via modified generalized barycentric coordinate interpolation of the map. The system we devised has facilities for contrast and feature enhancement, supports both regular and irregular grids, can deal with multi-field as well as multispectral data, and can produce heat maps, choropleth maps, and diagrams such as scatterplots.

Numerous methods have been described that allow the visualization of the data matrix. But all suffer from a common problem - observing the data points in the context of the attributes is either impossible or inaccurate. We describe a method that allows these types of comprehensive layouts. We achieve it by combining two similarity matrices typically used in isolation - the matrix encoding the similarity of the attributes and the matrix encoding the similarity of the data points. This combined matrix yields two of the four submatrices needed for a full multi-dimensional scaling type layout. The remaining two submatrices are obtained by creating a fused similarity matrix - one that measures the similarity of the data points with respect to the attributes, and vice versa. The resulting layout places the data objects in direct context of the attributes and hence we call it the data context map. It allows users to simultaneously appreciate (1) the similarity of data objects, (2) the similarity of attributes in the specific scope of the collection of data objects, and (3) the relationships of data objects with attributes and vice versa. The contextual layout also allows data regions to be segmented and labeled based on the locations of the attributes. This enables, for example, the map's application in selection tasks where users seek to identify one or more data objects that best fit a certain configuration of factors, using the map to visually balance the tradeoffs.

Contextual layouts preserve the context of the data with the associated attributes (variables). However, their linear mapping causes errors in the layout - similar data points and variable nodes may not map to similar regions, and vice versa. In this paper, we first unify the various data layout schemes and choose the Generalized Bary-centric Coordinates (GBC) plot as the standard way to describe them. Second, we propose three algorithms - distance spaced lay-out, iterative error reduction, and force directed adjustment - to reduce the layout error of variables to variables, data to variables and data to data, respectively. We find that the combination of these three algorithms can yield large improvements in the layout error and so achieve a more comprehensive layout. Third, we describe an interface, the GBC Error Explorer, which allows users to explore the error using a variety of visualization schemes combined with some interactions.

Nowadays, environment pollution is one of the most serious problems facing mankind, and more and more people are concerned about it. To solve it, we should get to know it first. In this paper, we present the method of tree-shaped multivariate visualization for hierarchical data, to reveal changes in the data and make comparisons among different pollutants and the time. The technique is based on a botanical tree metaphor. The resulting tree-like visualization can display many properties of the data and we can learn some useful information through the interactions. We also use traditional parallel coordinates as an auxiliary method and a complement to analyze the result.

A concept of shape coordinates system for visualization of data set is proposed in this paper. First, the visualization data set, visualization graphics, visualization process and visualization space are defined. Then, the definition, mapping, operation, theorems, properties and algorithms of shape coordinates system are described. Finally an example to visualize the data set as a tree shape graphics in the visualization space by using the shape coordinates system is given.