Big-Data Visualization

20 e've entered a data-driven era, in which data are continuously acquired for a variety of purposes. The ability to make timely decisions based on available data is crucial to business success, clinical treatments, cyber and national security, and disaster management. Additionally, the data generated from large-scale simulations, astronomical observatories, high-throughput experiments, or high-resolution sensors will help lead to new discoveries if scientists have adequate tools to extract knowledge from them. However, most data have become simply too large and often have too short a lifespan. Almost all fields of study and practice eventually will confront this big-data problem. Government agencies and large corporations are launching research programs to address big data's challenges. Visualization has proven effective for not only presenting essential information in vast amounts of data but also driving complex analyses. Big-data analytics and discovery present new research opportunities to the computer graphics and visualization community. This special issue highlights the latest advancements in solving the big-data problem via visual means, with four articles on new techniques, systems, or applications. In "Customizing Computational Methods for Visual Analytics with Big Data," Jaegul Choo and Haesun Park discuss the interplay between precision and convergence-two aspects that haven't received appropriate consideration in visual analyses so far. The authors propose customizing computational methods to include low-precision computation and iteration-level visualizations to ensure real-time visual analytics for big data. In "Feature Tracking and Visualization of the Madden-Julian Oscillation in Climate Simulation," Teng-Yok Lee and his colleagues present an integrated analysis and visualization framework for scientists to better understand the MaddenJulian oscillation (MJO) phenomenon from largescale spatiotemporal climate simulation data. The authors demonstrate how the tight integration of MJO domain knowledge, data analysis techniques such as feature tracking, and visualization methods such as Hovmoller diagrams and a virtual globe can lead to a powerful system for climate research. In "Visualizing Large, Heterogeneous Data in Hybrid-Reality Environments," Khairi Reda and his colleagues show how a new kind of visualization space called hybrid-reality environments can achieve scalable visualization of heterogeneous datasets. These environments synergize the capabilities of VR and high-resolution tiled LCD walls, letting users juxtapose 2D and 3D datasets and create hybrid 2D3D information spaces. The authors introduce two such environments-Cyber-Commons and CAVE2and some real-world applications. Finally, in "Exploring the Connectome: Petascale Volume Visualization of Microscopy Data Streams," Johanna Beyer and her colleagues describe a system for interactive exploration of petascale volume data of neural tissues generated by high-throughput electron microscopy imaging. This visualization-driven system lets users handle multiple volumes and incomplete data, restricts most computations to a small subset of the data, and achieves scalable computing with a multiresolution virtual memory. The authors applied the system to a mouse cortex volume with a resolution of 21,494 x 25,790 x 1,850 voxels.