CS: The Foundation of Modern Life.
Note: this was the feature article in Click! Magazine, 2015, volume II. Later that year, Dr. Jim Kurose, Assistant Director for NSF CISE, argued that booming undergraduate enrollments could be a "sea change" rather than a "wave" because “computing has infused a myriad of industry sectors and academic disciplines far beyond the tech sector,” taking note of CS @ ILLINOIS’s CS + X Program.
In what single University of Illinois department would you find faculty research ranging from construction site optimization, to social media algorithms, to bee genomes and evolution, to simulating how diseases spread? In CS @ ILLINOIS, of course—and for good reason.
Computer science underpins just about every aspect of modern life. The arts, science, business, medicine, and engineering all benefit from the computational power, modeling, and thinking found in computer science. And not surprisingly, CS @ ILLINOIS faculty are bringing their expertise to bear on a variety of vexing issues in a wide variety of fields.
One of the most costly issues related to major construction projects is the allocation of labor. “Right now, only about 50% of construction labor costs are productive,” said CS Associate Professor Derek Hoiem. “About half the time, labor is idle because of how hard it is to schedule where people should be.”
Construction managers are limited by how quickly they can reallocate their workers because most construction progress data is compiled manually. “Management can only respond to delays on a weekly basis, which ends up creating large delays over time that are very costly,” explained Hoiem, who is working with fellow Illinois engineering faculty on a novel autonomous vision-based robotic solution to monitor construction progress more quickly and cost effectively.
A computer vision expert, Hoiem is developing robust and rapid 3D reconstruction techniques to model the construction site based on images and video streaming from cameras strategically placed throughout the site by a quadcopter. He then aligns the 3D model with the planned model to automatically detect deviations.
“We can then identify parts of the construction that are behind schedule,” Hoiem said. “That would allow the managers to quickly reallocate resources in order to bring things up to speed.”
With funding from NSF, Hoiem and his colleagues—Civil Engineering Professor Mani Golparvar-Fard and Aerospace Engineering Professor Tim Bretl—are collecting data from a residence hall construction site on campus and at the new $500 million Sacramento Kings basketball stadium project in California.
According to Hoiem, the computer vision field has taken off in the last few years as techniques have become more robust and reliable. “Applications of the technology range from entertainment, to surveillance, to automated driving—anything where you want a machine to understand the world,” he said. “This is a very exciting time for computer vision.”
When Karrie Karahalios was growing up, she recalled her father climbing on the roof to adjust the TV antenna when a soccer game was on. “We didn’t understand electromagnetism or wave propagation, but we did see cause and effect—move the antenna a certain way and get a better picture,” said Karahalios, a CS associate professor.
Karahalios is taking the same approach to social media use. Through her research on human-computer interaction, she aims to inform people about algorithms social media sites use and how they shape people’s experiences—from what comes up in a Google or Yahoo search, to what Facebook users see in their news feed, to whether someone qualifies for a mortgage or auto loan. In essence, she’s on a quest to help people manage their interactions on the Web.
“Some people aren’t even aware of algorithms, and many people don’t know how to modify their behavior to get the algorithm reception they want,” said Karahalios, who recently published her research study on Facebook user habits and how they changed after people learned about the algorithms controlling their news feed.
In her study, Karahalios convened a demographically diverse group of Facebook users—nearly two thirds of whom did not know their news feeds were automatically edited. She ran each user’s Facebook feed through her FeedVis tool, which showed all the content posted by their friends.
“Our research was the first to discover that while companies assumed that people were aware of these algorithms, they were not,” Karahalios noted. “Our research was also the first to come up with some design practices to make people aware of them.”
People were shocked when they discovered that they’d missed posts from a family member or close friend, said Karahalios. Several months later, she followed up with the Facebook users from her study and learned that many of them had changed their behavior. For example, some users now used the Facebook settings to view their feed chronologically, some users lost trust in their feed and began visiting their close friends and family members’ Facebook pages directly to make sure they didn’t miss any important news.
“We have a right to know how these algorithms work, but they are locked behind proprietary walls,” she said. “People are coming up with complex strategies [to manage the algorithms], but the interfaces should support them.”
CS Associate Professor Saurabh Sinha has created a number of computational tools for the academic community. One of them is genomic analysis software that uses the concept of hidden Markov models to score the presence of biologically important patterns called transcription factor motifs in DNA.
Recently, he used the software to help an international team of researchers discover that gene regulation, or the process of turning genes on and off, underlies the evolution of social complexity in bees. This discovery, which was published in the June 5, 2015, issue of Science, provides new insights into the genetic changes that accompany the evolution of bee societies—something that scientists have tried to understand for more than a century.
Sinha and his colleagues, including Illinois Entomology Professor Gene Robinson, used software to analyze the genomes of 10 different bee species, identifying biologically meaningful patterns in their DNA. According to Sinha, an increase in the presence of a particular pattern near a gene indicated a more socially complex bee colony.
During the last 10 years, Sinha has made other important contributions to understanding the gene regulatory process by working with Robinson and other faculty at the Carl Woese Institute for Genomic Biology (IGB) on the Urbana campus. For example, Sinha conducted a computational analysis of gene regulatory relationships in three different species (house mouse, stickleback fish, and honey bee), enabling his fellow IGB colleagues to conclude that animals not only share genes for some common physical traits but may also share a genetic toolkit for behavioral traits as well.
According to Sinha, the merging of computer science and other disciplines is a welcome change, due in part, to well-trained computer scientists moving out of their core research area and learning about a very different domain of inquiry.
“The attitude towards interdisciplinary work has started changing from one of ‘send me your data and questions, and I will build a method to solve them,’ to the much more effective attitude of ‘let’s together find out what the questions are and how computing can help answer them,’” he said. “In other words, the big change is in the minds of people.”
Working with lead researchers at Virginia Tech University, CS Professor Laxmikant “Sanjay” Kale is helping discover how contagions like the H1N1 and Ebola viruses propagate through populations. Their EpiSimdemics project simulates the spread of an epidemic in extremely large and realistic social contact networks, capturing dynamics among co-evolving entities.
In addition, their simulation model can analyze what-if scenarios, such as predicting the effect that closing schools or broadcasting precautions to the public would have on the spread of the disease.
“This could be a great tool in the hands of public health officials,” said Kale, noting how an increasingly urban and mobile population enhances the chances of a future worldwide pandemic.
Solving a complex problem like this requires serious computing power, and, more importantly, sophisticated parallel applications capable of exploiting that power. Kale recognized that automating resource management is a key to simplifying application development, which led to the design of the Charm++ adaptive runtime system. Charm++ is available via the web so other scientists can develop new applications for large-scale simulations.
Kale started developing Charm++, which is based on the C++ programming language, nearly 20 years ago, in order to make it easier to develop software for supercomputers, while enabling efficient utilization of the machines. He and his research group have continued to develop, improve, and evolve the system based on feedback from collaborative application development.
“We have been working a long time to develop a really powerful parallel programming system,” Kale said. “Charm++ is one of the few academic software systems that has had impact and sees actual use in applications on supercomputers.”
One such application was a collaboration with Illinois Biophysics Professor Klaus Schulten several years ago in developing Nanoscale Molecular Dynamics (NAMD) software. Recently, NAMD was used by Schulten and collaborators in the world’s first simulation of the precise chemical structure of the HIV capsid. Kale’s simulation code accounted for the interactions of 64 million atoms.
According to Kale, more than 70,000 researchers have used NAMD to conduct a variety of biophysical studies. Kale’s group has created other scientific applications built with the Charm++ system that enable astronomers to study the origins of the universe (ChaNGa) and engineers to study quantum-mechanical details of photovoltaic materials (OpenAtom), which may lead to better solar cells.
“We are not experts in quantum chemistry or astronomy or epidemiology,” he said. “Our expertise is in parallel programming and how to make parallel computing easy and efficient for scientists and engineers who want to develop these kinds of models.”