3 NCSA interns honored for progress on projects

03.12.15 -

By Elizabeth Murray

The 26 U of I undergraduates in NCSA’s SPIN (Students Pushing Innovation) internship program presented lightning talks on the progress of their projects at the close of the fall semester. A panel of campus judges honored three for the significant work they have done so far.

Matthew Ho: Understanding human behavior

Matthew Ho is a physics major working with mentor Guy Garnett (Illinois Informatics Institute; associate professor of Music) to develop a mathematical protocol for defining human expression in movement. After first determining how humans identify human expression—e.g. the difference between angry and calm movements—in a clear and comprehensive way, the group then hopes to translate that understanding over to computers, so that they may similarly identify them.

“What we are trying to achieve is a very significant, not to mention difficult, step in both our understanding of human behavior but in our capacity to define and predict nature through computer science,” says Ho. “In the future, we may finally be able to perfect extremely complex concepts, such as human communication and artificial intelligence.”

Garnett says Ho has been a key addition to his team. “His experience in physics has provided a crucial link between the machine learning group and the understanding and representation of the physical component of movement.”

The work also has the potential to be very influential in the field of art and dance. Ho believes it opens up a new frontier for artists to use computers and technology to design and create new works of expression. The team has begun writing papers on their findings from the fall semester.

Chuck Rozhon: Streamlining scientific data

Chuck Rozhon is a computer science major working with Matthew Turk (NCSA research scientist) on GPU-based volume rendering, specifically for the yt project—Turk’s python package for analyzing and visualizing volumetric, multi-resolution data from astrophysical simulations and radio telescopes with a burgeoning interdisciplinary community.

Currently yt uses a CPU-based rendering algorithm, which works, but can also be very slow. As Rozhon points out, the process for volume rendering is inherently very parallel, and graphics processors are particularly good at doing parallel operations, making yt a great candidate for GPU-based rendering.

Turks says his work with Rozhon has been mutually beneficial. “I feel like I'm learning things from him all the time,” he says. “From the beginning, he simply dived into the work, bringing with him creativity, excitement, and an incredible eagerness.”

Rozhon believes that his work to add GPU-based rendering capability will make it easier and faster for scientists and users of yt to get what they desire out of a 3D data-set. “It will allow them to customize how they want their data to look in an intuitive way, allowing the scientists who use yt to further their research,” he explains.

Nate Russell: Uncovering biological insights

Nate Russell, a returning SPIN student and industrial and enterprise systems engineering major, is working with Colleen Bushell (Applied Research Institute, NCSA affiliate) on developing semi-supervised learning algorithms for feature selection and ranking in Multi-Omics data-sets.

Russell says he and his group want to uncover patterns in very high-dimensional data where there are very few samples. “We would like to relate biological features to health outcomes like disease state or toxicity,” he explains. “We strive to produce visualizations that prompt underlying questions and enable deeper biological insight.”

Bushell says that Russell’s energy and optimism have added a spark to both the group and the research. “He is intense about his interests and goals and thrives on discussing ideas and solving difficult problems,” she says.

Russell believes the group’s latest computational methods and visualizations have already had societal impact. The method has been used in nearly a dozen different data-sets that range from animal nutrition to pelvic pain to breast cancer.

Next, he would like to explore unsupervised algorithms and fuzzy classifiers that support hypothesis generation for researchers and clinicians. “My personal hope is that advanced analytics tools like ours become commonplace in biology and help advance the field,” concludes Russell.

For more about all of the 2014-2015 SPIN projects, visit http://spin.ncsa.illinois.edu/interns/.

Recruiting for 2015-2016 SPIN students will begin at an open house from 3 to 5 pm April 8 at the NCSA Building. All University of Illinois undergraduates are invited to attend to learn more about the projects and opportunities at NCSA!