Yunzhe Wang

ModulGraph: modularity-based visualization of massive graphs

Large graph visualization has become a dominant problem in multiple big data analytics domains, including media analytics, social network dynamics, resource management in cloud computing environments, air travel, large networks. A practical approach to displaying massive graphs is by partitioning according to well defined domain-dependent attributes. However, graph visualization in the presence of incomplete information is an open challenge in many applications. In order to better visualize and understand patterns in large graphs, local pattern discovery becomes a critical step in deciding the structural components of graph visualization. In this paper, we present a modularity-based graph visualization method, termed as the ModulGraph. The ModulGraph is a hierarchical representation that treats a graph as a set of modules. The main objective of this work is to hierarchically detect graph patterns in order to visualize large graph data and adapt the interconnecting structures to potential interactions between local module streams. Our main contribution is a graph visualization method that can flexibly detect the local patterns or substructures, called modules, in large graphs. The second contribution is a hybrid modularity measure. This measures hierarchically the cohesion of the graph at various levels of details. We aggregate clusters of nodes and edges into several modules for the purpose of reducing the overlap on the display. Graph patterns of modules are processed by the ModulGraph system in order to avoid information loss while a sub-graph is represented as a single node. Our experiments show that this method can support large-scale graph visualization for visual media exploration and analysis.

Cloudet: A Cloud-Driven Visual Cognition of Large Streaming Data

Streaming data cognition has become a dominant problem in interactive visual analytics for event detection, meteorology, cosmology, security, and smart city applications. In order to interact with streaming data patterns in an elastic cloud environment, we present a new elastic framework for big data visual analytics in the cloud, the Cloudet. The Cloudet is a self-adaptive cloud-based platform that treats both data and compute nodes as elastic objects. The main objective is to readily achieve the scalability and elasticity of cloud computing platforms in order to process large streaming data and adapt to potential interactions between data stream features. Our main contributions include a robust cloud-based framework called the Cloudet. This is a cloud profile manager that attempts to optimize resource parameters in order to achieve expressivity, scalability, reliability, and the proper aggregation of the compute nodes and data streams into several density maps for the purpose of dynamic visualization.

Cloudets: Cloud-based cognition for large streaming data

Big data cognition has become a dominant problem in interactive visual analytics for event detection and response, metereology, cosmology, and large smart city applications including traffic monitoring and management, search and rescue operations, crowd management and logistics. The main problems are mainly due to big data volume and velocity and, in some cases, variety in both dimension and type. A practical approach to understanding and viewing big data features is through streaming operations. Streaming allows for both volume and velocity characteristics of big data, and often, for variety as well. However, performing analytics at interactive rates is currently an open challenge in most big data applications. Cloud computing platforms provide practical support and leverage to solving some of the big data and visual analytics problems, especially when dealing with the volume and velocity characteristics of current data generation. In order to interact with streaming data patterns in an elastic cloud environment, we present a new elastic framework for big data visual analytics in the cloud, the Cloudet. The Cloudet is a self-adaptive cloud-based platform that treats both data and compute nodes as elastic objects. The main objective is to readily achieve the scalability and elasticity of cloud computing platforms in order to process large streaming data and adapt to potential interactions between data stream features. Our main contributions include a robust cloud-based framework, the Cloudet, which can flexibly process the streaming data and applications to illustrate the setup and operations of this framework. The framework includes a cloud profile manager that attempts to optimize the cloudet parameters in order to achieve expressivity, scalability, reliability, and the proper aggregation of the data streams into several density maps for the purpose of dynamic visualization of data features.

Cognitive Visualization of Popular Regions Discovered From Geo-Tagged Social Media Data

This article focuses on the cognitive exploration of photo sharing data which contain information about the location where the photo was taken and potentially some description about the photo. Therefore, the features of photo-spots can be deduced. Spots with similar features constitute a region of cognitive interest. The objective is to identify these regions and allow users to explore into regions of interest by cognitive understanding of their features. The authors propose an approach that makes use of semantic analysis, data clustering, and cognitive visualization. In this article, the authors introduce the design of an interactive visualization interface which projects photo sharing data to cognitive social activity map components. The contributions are two-fold. First, the authors put forward a novel social-media data classification method. Second, the authors suggest a new method to explore social activity maps by discovering regions of cognitive interest. Experiments are performed on the Flickr dataset.

Smooth animation of structure evolution in time-varying graphs with pattern matching

Drawing a large graph into the limited display space often raises visual clutter and overlapping problems. The complex structure hinders the exploration of significant patterns of connections. For time-varying graphs, it is difficult to reveal the evolution of structures. In this paper, we group nodes and links into partitions, where objects within a partition are more closely related. Besides, partitions maintain stable across time steps. The complex structure of a partition is simplified by mapping to a pattern and the evolution is exposed by comparing patterns of two consecutive time steps. We created various visual designs to present different scenarios of changes. In order to achieve a smooth animation of time-varying graphs, we extract the graph layout at each time step from a super-layout which is based on the super-graph and super-community. The effectiveness of our approach is verified with two datasets, one is a synthetic dataset, and the other is the DBLP dataset.

Cognitive exploration of regions through analyzing geo-tagged social media data

Social media has now become a pervasive global communication channel. Many applications and platforms have become available for users to post messages, follow friends and share experiences. Due to the high frequency with which users update their states, a large amount of data is being generated around the world every second. By analyzing this data, valuable patterns can be extracted such as the distribution of users, their common interests, activities, locations visited, etc. In this paper, we focus on the cognitive exploration of photo sharing data. Traditionally, each photo sharing record comes with information about the location where the photo was taken, a timestamp, and potentially some description about the photo. Therefore, we can often deduce the features of photo-spots. Spots with similar features constitute a region of cognitive interest. The primary goal of this paper is to identify these regions and allow users to explore into regions of interest by cognitive understanding of their features and patterns of feature propagation in time. To achieve this goal, we propose an approach that makes use of semantic analysis in big data sets, data clustering, and cognitive visualization design issues. Our contributions are two-fold. First, we put forward a novel social-media data classification method. This is based on cognitive semantic analysis. Second, we suggest a new method to explore social activity maps by discovering regions of cognitive interest. In this paper, we introduce the design of an interactive visualization interface which projects photo sharing data to cognitive social activity map components. Experiments are performed on the Flickr dataset.

Module-based visualization of large-scale graph network data

AbstractThe efficient visualization of dynamic network structures has become a dominant problem in many big data applications, such as large network analytics, traffic management, resource allocation graphs, logistics, social networks, and large document repositories. In this paper, we present a large-graph visualization system called ModuleGraph. ModuleGraph is a scalable representation of graph structures by treating a graph as a set of modules. The main objectives are: (1) to detect graph patterns in the visualization of large-graph data, and (2) to emphasize the interconnecting structures to detect potential interactions between local modules. Our first contribution is a hybrid modularity measure. This measure partitions the cohesion of the graph at various levels of details. We aggregate clusters of nodes and edges into several modules to reduce the overlap between graph components on a 2D display. Our second contribution is a k-clustering method that can flexibly detect the local patterns or substructures in modules. Patterns of modules are preserved by the ModuleGraph system to avoid information loss, while sub-graphs are clustered as a single node. Our experiments show that this method can efficiently support large-scale social and spatial network visualization.Graphical AbstractGraphical Abstract text

Fast content-aware resizing of multi-layer information visualization via adaptive triangulation

Abstract Visual graphics and image-based content have become the pervasive modes of interaction with the digital information flow. With the immense proliferation of display systems and devices, visual content representation has become increasingly challenging. Classical static image resizing algorithms are not directly suitable for the current dynamic information visualization of streaming data flows and processes because most of the visual content often consists of superimposed, multi-layered, multi-scale structure. In this paper, we propose a new adaptive method for content-aware resizing of visual information flow. Scaling is performed by deforming a hierarchical triangle mesh that matches the visual saliency map (VSM) of the streaming data. The VSM is generated automatically based on a series of predefined rules operating on a triangular mesh representation of visual features. We present a linear energy function to minimize distortions of the triangular deformations to perceptually preserve informative content. Through multiple experiments on real datasets, we show that the method has both high performance as well as high robustness in the presence of large differences in the visual aspect ratios between target displays.

Cognitive visual analytics of multi-dimensional cloud system monitoring data

Hardware virtualization has enabled large scale computational service delivery models with significant cost leverage and has improved resource utilization of cloud computing platforms. This has completely changed the landscape of computing in the last decade. It has enabled very large-scale data analytics through distributed, high performance computing. However, due to the infrastructure complexity, end-users and administrators of cloud platforms can rarely obtain a complete picture of the state of cloud computing systems and data centers. Recent monitoring tools enable users to obtain large amounts of data with respect to many utilization parameters of cloud platforms. However, they often fall short of maximizing the overall insight into the resource utilization dynamics of cloud platforms. Furthermore, existing tools make it difficult to observe large scale patterns making it difficult to learn from the past behavior of cloud system dynamics. New operating platforms for cloud management and service provisioning allow live migration and dynamic resource re-allocation at multiple levels of the hardware virtualization layers. Hence, it has become necessary to provide cognitive visualizing tools for monitoring the activities in an active cloud environment. In this work, we describe a perceptual-based interactive visualization platform that gives users and administrators a cognitive view of cloud computing system dynamics. We define machine states and aggregate states at multiple levels of detail to construct a multiview presentation of the resource utilization according to the scalability and the elasticity features of a cloud computing system.

VisQuery: Visual querying of streaming data via pattern matching

Querying streaming data is becoming a dominant problem in big data analytics. A practical approach to querying streaming data is through traditional databases that have been modified to support streams, such as MySQL. However, conditional selection for querying data streams is currently an open challenge. We present a new visual framework that provides a more intuitive querying interaction for streaming data by combining visual selections on patterns with image processing techniques in order to better identify regions of interest. The main contribution of this paper is a novel method for matching patterns among normalized frames via feature vector clustering.