Story Tips From TACC

To arrange for an interview with a researcher, please contact the External Relations Manager. For more information on TACC, its resources, and its research and development activities, and for general media-related questions or comments, contact our Public Relations Coordinator, Faith Singer-Villalobos.

Big Data

Just a few generations ago, storing information was costly and usually reserved for only the most critical data (e.g. astronomical observations, census calculations, commodity prices). In the past few years, thanks to the lowering cost of data storage and the growing number of platforms for gathering data, we are facing a new challenge: making sense of the vast amounts of data that we're able to connect.

How is this data — especially scientific data — being stored, mined, managed and used? TACC experts in data mining and data-intensive computing can provide insights into today's big data challenges and ways to address them. We have compelling case studies ranging from the National Archives to smart grid projects to archeological digs.

Super-Resolution by Example

Dr. Jason Lawrence, assistant professor of Computer Science at the University of Virginia, is a lead researcher on an NSF-funded project that is investigating a radically different approach to image processing tasks. Dr. Lawrence's team poses the question: can image quality be improved using information from millions of on- line images?

The team is working with approximately 10 million high-resolution digital images scraped from Flickr to develop a system that can recover plausible details. The key barrier in deploying this type of example-based image enhancement system is constructing methods for efficiently searching this large amount of image data to identify similar patches.

"[The Ranger supercomputer] has proved to be an enabling technology that has carried our work from the proverbial drawing board to reality," Lawrence says. "At the current date, we have successfully performed nearest neighbors queries over 10^12 training patches in ~80 minutes using 4,096 processing cores. The same feat would require roughly ~228 days of processing on a single computer!"

Arietta, S., Lawrence, J., Building and Using a Database of One Trillion Natural-Image Patches, IEEE Computer Graphics and Applications, 31(1), pp. 9--19. Jan/Feb 2011.

Exploring New Methods for Large-scale Graphene Production

Researchers and industry are learning more about the excellent properties of graphene everyday. However, established manufacturing methods have not proven ideal for the large-scale creation of graphene sheets.

An alternative manufacturing method involves the reduction of graphene oxide, a material that shares the same atomically thin structure as graphene. A team of researchers, including Vivek B. Shenoy, professor of Engineering at Brown University, is using the Ranger supercomputer to perform molecular dynamics simulations in order to study the atomistic structure of graphene oxide as it is reduced.

By elucidating the chemical changes in the substance, they are helping to find a new ways to produce a potentially transformative material.

Can We Trust Computer Models?

UT researchers aim to quantify our uncertainty in important areas like climate modeling and defense

We are now relying on computer models to predict events of enormous importance to our welfare and security, such as climate change, the performance of energy and defense systems, the biology of diseases, and the outcome of medical procedures.

Just how good are our predictions of such complex phenomena, and how can we quantify the inevitable uncertainty that arises when making such predictions?

Several teams of researchers at the University of Texas at Austin are exploring this question. Using the power of supercomputers to study climate change, space shuttle reentry, and computer-aided surgery, these projects are adding additional layers of information to predictions and helping researchers andpolicy makers understand how much they can trust computer models.

Patient-specific Drug Delivery

A vast majority of heart attacks occur when there is a sudden rupture in the coronary arteries, exposing its core materials to the blood flow, and forming blood clots that can cause deadly blockages. The diseased arteries can be treated with drugs delivered locally to these rupture-prone plaques termed "vulnerable plaques". In designing these local drug delivery devices, important issues regarding drug distribution and targeting need to be addressed to ensure device design optimization for maximum therapeutic efficacy.

A computational tool-set was developed to support the design and analysis of a catheter-based local drug delivery system that uses nanoparticles as drug carriers to treat vulnerable plaques and diffuse atherosclerosis. Simulations were run on a 3D patient-specific multilayered diseased coronary artery segment obtained directly from CT-imaging data and the effect of artery wall and plaque inhomogeneity on drug distribution was analyzed.

The tool is now poised to be used in medical device industry to address important design questions such as, "given a particular desired drug-tissue concentration in a specific patient, what would be the optimum location, particle release mechanism, drug release rate, drug properties, and so forth, for maximum efficacy?"