Interdisciplinary research institute established in new hub of Lehigh innovation
The topic of interdisciplinary research must be pretty timely if the journal Nature dedicates a special issue to it, as it recently did. In an opinion piece, the editors outlined the impracticality of attacking the globe’s most pressing problems from scientific departments encased in silos. “The best interdisciplinary science comes from the realization that there are pressing questions or problems that cannot be adequately addressed by people from just one discipline,” the editors wrote. “An interdisciplinary approach should drive people to ask questions and solve problems that have never come up before.”
Lehigh is responding to the demand for innovative approaches to scientific inquiry with the creation of three inaugural Interdisciplinary Research Institutes (IRIs) designed to bring together top scholars across the university and their external partners to pursue large-scale, transformative research. The institutes are the fruit of a faculty-led Envisioning Process, which was among the first initiatives launched by Steve DeWeerth, professor and dean of P.C. Rossin College of Engineering and Applied Science. The process tasked faculty from Rossin and the other Lehigh colleges with evaluating and identifying strategic strengths, opportunities for growth, and specific interdisciplinary themes for which Lehigh is uniquely positioned to assume a global leadership role.
The envisioning team – looking outward to find the most vital and exciting research areas and inward to mine Lehigh’s native strengths – identified the themes for the inaugural Institutes: the Institute for Functional Materials and Devices (I-FMD); the Institute for Data, Intelligent Systems, and Computation (I-DISC); and the Institute for Cyber–Physical Infrastructure and Energy (I-CPIE).
Each Institute and its faculty will incubate and catalyze big ideas, will pursue large-scale extramural funding, and will promote the impact and visibility of Lehigh research to the world. The University has committed to providing interdisciplinary facilities, “venture” funding, and Institute staffing to support the success of these endeavors. In addition, the Rossin College is supporting the growth of these Institutes through a commitment to strategic faculty hiring to enhance and extend Lehigh research in these areas.
The new Institutes will position Lehigh as a thought leader and destination across these themes. Hector Muñoz-Avila, I-DISC co-director and Professor of Computer Science and Engineering, also foresees that the Institutes’ portfolios will dovetail with each other. “As an example,” he says, “the work of I-DISC will have broad areas for collaboration with I-CPIE and I-FMD, in the context of the increasingly-crucial role that data and computation plays in those domains.”
I-DISC’s domain will extend across intelligent systems, data, computation, optimization, robotics, machine learning, and related fields. “Our team includes faculty from across Rossin College, as well as government relations, biological sciences, and economics. We identified data, intelligent systems, and computation as places where Lehigh already had great strength that could be put in a position to do more integrated research and have a significant impact,” says Muñoz-Avila. “We have seen in recent years a data revolution in terms of the kind and amount of data that is available, and with advances like the Internet of Things there will only be more.”
The Institute has a running start in part because it was preceded by Lehigh’s Data X initiative, a cross-curricular program that infused interdisciplinary, data-centric research throughout the university. Part of Data X’s mandate was a significant number of faculty hires, and as the Institute gears up, more are in the pipeline. “Lehigh has committed significant resources to support the research institutes, and will be doing more faculty hiring in strategic areas such as data analytics, cyber–physical systems, and robotics, which will augment the numerous Data X hires over the past few years,” says DeWeerth.
The fundamental demand of the research institutes is that they be foundational in their structure as well as their deliverables. “I-DISC will create bridges across the university and be a fertile space for collaborative work, and the research we do will attack at the root some of the most pressing problems in technology and society,” says Katya Scheinberg, I-DISC co-director and Harvey E. Wagner Professor of Industrial and Systems Engineering. “It’s meant to be for the whole university, and we expect that the research it produces will generate solutions with broad applications. It may be engineering-focused, but it won’t be engineering-centric.”
Existing partnerships with Lehigh’s College of Business and Economics (CBE) provide an example of how such cross-pollination creates benefits. “We already have two specific people who were hired in consultation with the computer science department, and we have made other hires of people who combine business and data and computation in their research and teaching areas,” says Georgette Chapman Phillips, the Kevin and Lisa Clayton Dean of the College of Business and Economics. “Our relationship with computer science gives our students the opportunity to have high levels of expertise in data science in addition to business. From a research perspective, I-DISC will push our faculty to the forefront in this area. This field is a perfect example of how Lehigh works well across colleges.”
The environment for collaboration doesn’t happen by itself, states a 2016 report on interdisciplinarity commissioned by the Global Research Council. The study, in outlining the increasingly vital role that interdisciplinary research will play in resolving society’s grand challenges, specifically detailed that creating conditions for collaboration was of primary importance to the success of interdisciplinary research programs. “Much of the literature argues that funders should not assume that the conditions required for interdisciplinarity can happen naturally without proactive support,” reads the report. “Instead, consideration should be given to the practical steps and mechanisms necessary to foster and support research across disciplinary boundaries.”
The newly renovated Building C on the Mountaintop Campus, which will house the Institute, was in part specifically designed to be a catalyst for just that.
A (r)evolutionary overhaul
Built both for comfort and speed, Building C has broad, clear sight lines, glass and steel encased classrooms and offices, and panoramic views of the Lehigh Valley; it is also wired with up-to-the-minute data and networking connectivity. There are plenty of welcoming meeting places where you might run into faculty from other fields, and have the kind of impromptu conversations that lead to unexpected insights. The building sports a feature called Mixing Boxes, glass-enclosed meeting rooms that overlook the expansive bays. It’s not all glam though. In a reverent nod to the manufacturing history of the former Bethlehem Steel structure, architects preserved many of the original features of the building, including the shell which houses the more contemporary internal fixtures.
If the walls could speak, they would say that this is a place where work gets done. Lab spaces and bays will facilitate the kind of experiential learning that Lehigh is known for and that students demand, and the overall feeling of Building C is an agreeable balance of gleam and grit, an homage to an industrial past, and an eyes-forward view of the research to come.
Dan Lopresti, Professor and Chair of the Department of Computer Science and Engineering, was involved in many of the discussions about the design of Building C. “I’m lucky in that I get to travel around the world and see a lot of really cool buildings, and they all have aspects that are quite impressive. Building C has the flavor of buildings at Facebook and Google, and that is great for our students, but it is really unique,” he says. “It’s not like anywhere else you’ve been, and when you step inside you feel that right away.”
Scheinberg’s work is a prime example of the type of inquiry with broad application that the Institute hopes to foster. Scheinberg focuses on optimization, the use of sophisticated mathematical tools to generate efficient solutions. “Optimization is extant in all engineering fields and across all data problems, which is why it is foundational and why we are good partners in the Institute. It is also at the core of most machine learning methods, which is a very important contemporary field of study,” says Scheinberg.
Scheinberg and her colleagues have received a Transdisciplinary Research in Principles of Data Science (TRIPODS) grant from the National Science Foundation, a three-year, $1.5 million award. The TRIPODS team includes Scheinberg as principal investigator, and co-investigators Frank E. Curtis, and Martin Takáč, associate professor and assistant professors, respectively, in the department of industrial and systems engineering, as well as Han Liu of Northwestern University and Francesco Orabona of SUNY-Stony Brook. The grant comprises three branches of study— theoretical computer science, statistics and applied mathematics—and a goal of the project is to maximize research potential by bringing together researchers from the three different communities. “You can have people in these three fields basically working on the same problem, but using different terminology, going to different conferences, and publishing in different journals,” says Curtis. “It’s a very inefficient process. We should be sharing our expertise, and that is one of the things we hope to accomplish.”
At the core of most machine learning problems is an optimization problem, Curtis explains. “If you want to get a computer to recognize a picture of a dog, one of the ways we do that is to get a bunch of images that have been classified or labeled, and give the machine enough images that it will learn what the essence of a dog is. Humans do that naturally, since we look at pictures holistically. A computer reads images zeroes and ones. There are large scale issues, but in essence, it is a computer operation where given an input, you want generate an appropriate label. That is where optimization comes in. You want the best classifier of all this data, and when you say ‘best,’ you are talking about optimization and the use of mathematical processes to get that.”
Inspired by neural networks modeled on the brain, deep learning techniques are making things like self-driving cars a reality. While these techniques have been around for decades, the deep learning revolution in recent years has been made possible by today's increased computational firepower. However, there remain enigmas embedded in deep learning tools.
"Because of the complexity of these neural networks, we don't always know why they give the outputs that they do," says Curtis. “The decision-making process is often opaque to developers, and in the case of a self-driving car, where incorrect decisions can be catastrophic, more transparency is required. These problems are so large that they demand experts from all fields working together to make progress.”
Miaomiao Zhang, assistant professor of computer science and engineering and a recent Data X hire, is pursuing new imaging techniques to improve the diagnosis and treatment of brain diseases. One of her current projects is extracting critical data information from caches of MRI (magnetic resonance imaging) scans to advance the understanding of brain disorders, and also make it possible to predict who might be vulnerable to neurodegenerative diseases such as Alzheimer’s.
Zhang is developing machine learning algorithms that analyze the brain shape variability across large populations, hence differentiating healthy groups and those affected by neurodegenerative diseases. Brain degeneration in maladies like Alzheimer’s disease can involve the loss of certain brain structures, changes in shape or shrinkage of the brain.
“The brain structure is extremely complex obviously, which makes this analysis challenging,” says Zhang. “The imaging data that fully captures the details of human brains lives in a high-dimensional space, and we need to work with millions of unknown parameters simultaneously.”
“Right now, the clinicians can detect a few areas of the brain associated with the disease, but it isn’t clear whether these are the only structures that matter. Since there may be other affected areas, we are looking into every detail of the entire brain to determine if there are other regions that possibly relate to the disease.”
Zhang’s work is complicated by two major factors. First, given the intricate morphology of the brain, MRI images are difficult to analyze. Second, the high volume of data in a full MRI of the brain contains a stupefying amount of information, a computational problem that demands simplification. In addition, MRIs can contain a lot of noise; Zhang’s algorithm will extract, to the degree possible, only the useful and relevant data. Moreover, since MRIs are not particularly common, Zhang’s work invokes what is known as the Curse of Dimensionality.
“We have very detailed MRI scans, but a relatively small population to analyze for this type of study,” Zhang says. “As a result, we have to find ways to make the image analysis as precise as possible to overcome that. Our work is ongoing, but it’s promising for the future.”
Zhang thinks the upcoming launch of Lehigh’s College of Health will boost her work along with medical research all over campus. “The involvement of clinicians and surgeons will provide great opportunities for multidisciplinary research collaborations among computer scientists like me along with bioengineers, cognitive psychologists, and others to solve real clinical problems.”
Robotics will be another vital spoke in the wheel of research at the Institute, and the soaring, 60-foot high bays will provide enough space for a small air force of drones, or for other uses yet undreamed of. John Spletzer, associate professor of computer science and engineering and head of the VADER robotics laboratory, and Joachim Grenestedt, professor of mechanical engineering and mechanics and director of Lehigh’s Composites Lab, have collaborated on groundbreaking projects using automated operations in the past, and are doing so again with watercraft.
Dubbed the Lehigh Ocean Research Craft Autonomous (LORCA), the LORCA boats are unmanned capsules about the size of a small porpoise and equally streamlined, with a top speed that a porpoise might envy. The automated craft, uploaded with map data, will hit 50 mph on a straightaway. Self-righting, the craft can be tossed into the water off a pier, and the composite shells are tough enough to withstand ocean waves. The self-driving drones could be used for surveillance, security, rescues or mapping, says Grenestedt.
“The ocean bottom can move a lot with storms. With the increase in the severity of hurricanes, it is a dynamic environment,” he says. “The LORCA craft fitted with echolocation equipment could replace the tedious manual mapping process that is now used, where divers literally plunge long sticks to the sea floor to take readings.”
Privacy is one of the hottest issues in data science, and a research focus of Haiyan Jia, assistant professor in the department of journalism and communication within the College of Arts and Sciences, and a Data X hire. “I teach data journalism, digital media and privacy in my classes. Our field has had a lot of disruptions recently, and so the discussions with students are pretty lively these days,” she reports. “I think Lehigh was ahead of its time in including data science as part of the journalism and communication curriculum.”
The variety of human behaviors that affect privacy issues play a large role in Jia’s research. “Data technology is creating so many benefits that there is no chance of reversing its influence, so we are left with the question of how to design a better future that minimizes risks,” Jia says. “People are in an active role online; they aren’t passive consumers of media. They may begin to use social media to connect with friends, but also to create their own personal identity, check up on news, or see what other people are doing, in other words, surveillance.”
Understanding the matrix of motivations and behaviors is integral to finding privacy solutions, which, humans being human, is no simple task. “One of the things I focus on is the social aspect to the problem. We typically think of privacy in terms of a person, but privacy exists in a social context, and people make decisions as part of a larger community.”
Another asset for Jia’s work will be a digital media lab that is being installed in Building C. “My research is empirical and experimental, and would benefit greatly from a lab for this kind of research,” she says. “Dan Lopresti and Jack Lule [chair of the department of journalism and communication] are both very supportive of creating a space for this empirical, interdisciplinary work. The university has been very supportive, and we hope that the lab will be a place for faculty to conduct research, and for students to do projects as well.”
All told, there could scarcely be a more auspicious moment for a new institute dedicated to collaborative work in data science. “Demand is through the roof for data analytics and computation among the students in our college,” Dean Phillips reports. “And frankly, it is a functional necessity for our students.”
“This societal importance of this field can’t be underestimated,” says DeWeerth, “and it is a field where Lehigh can and will be a leader.”