Weekly Bulletin
November 15, 1999
Vol. 89, No. 9
[<] [>] [archive]


Thinking about math
Real-world problems
Randall manages musical banquet
In print
Nassau Notes
Page one
In the news

Real-world problems

PICASSO initiative fosters interdisciplinary research related to computer science

By Steven Schultz

When Jaswinder Singh was a graduate student at Stanford, his adviser encouraged his choice of research but warned him that it was a risky one.

Singh spent many hours in the astrophysics department struggling with something called the N-body problem. The trouble was that his field wasn't astrophysics; it was computer science.

His idea was that computer systems design and the many disciplines that rely on computing should not insulate themselves from each other. "I see computers as a means to solving real-world problems, and real-world problems as the drivers we must use in designing hardware and software," says Singh. "That interdisciplinary style was not always rewarded, because of the difficulty of evaluating it within traditional categories.

"Fortunately," he says, "that has changed."

Now an associate professor of computer science at Princeton, Singh is directing a new effort that aims to foster interdisciplinary education and research related to computer science. The initiative, called PICASSO, is intended to train researchers who form bridges between computer science itself and other disciplines that increasingly use powerful computers as tools for basic research or applied information processing.

Fundamental changes

"In science, fundamental changes are occurring in the way research is conducted," says Singh. The two traditional modes of sciencetheory and experimentationhave in recent years been joined by a third approach called computational science. In this emerging field, scientists use computer simulations, rather than conventional experiments, to tackle complex problems.

Singh, for example, is collaborating with molecular biologists to help determine the structure of proteins and to develop a computer model of the immune system. High-performance computing is also playing a major role in other areas, such as engineering, finance and operations.

"There are vast numbers of problems that can be addressed by computational techniques and that you couldn't even think about solving 10 years ago," says University provost Jeremiah Ostriker, principal investigator of the grant that supports PICASSO. "This program provides a coherent structure for our graduate science education that really allows for interdisciplinary work."

The PICASSO program is expected to generate benefits in three directions. First, researchers in the various disciplines will be better equipped to take advantage of new developments in high-performance computing, such as parallel computers and novel problem-solving methods.

Second, computer scientists will gain a greater understanding of the aplications of computers in other disciplines. "This will enable them to design even better methods and systems, thus leading to major advances in computer science itself," Singh says. "Increasingly, computer science needs to strengthen its outward-looking aspects."

As these two things happen, Singh hopes the third benefit will follow: a new brand of truly interdisciplinary researchers will cement the bridges between disciplines, and entirely new research areas will crop up at the boundaries between existing ones.

PICASSO is also expected to address the exploding number of information processing and Internet services, many of which requre new methods and technologies. These areas are sure to drive interesting new research in computer science and will require a high level of collaboration between users and designers, says Singh.

Computation pipeline

A unique aspect of the new program, compared to traditional computational science programs, is that it will provide integrated research and training in the entire computation pipeline rather than focusing on a small subset of it, according to Singh.

The stages of this pipeline include: developing the mathematical model that describes the problem; creating a method, or algorithm, for solving the model; putting that method to use on parallel hardware, and perhaps designing new hardware and software systems; and creating innovative ways to analyze and display the results. In the last area, students may work closely with the new room-sized display wall project in computer science being led by computer science professor Kai Li, a member of the program's executive committee.

Ostriker noted that PICASSO builds on Princeton's traditional strengths in computational as well as computer sciences, which go back to physicist John von Neumann, who made many seminal contributions to computing, and astrophysicist Martin Schwarzschild who used the first computers in the 1950s to model the structure and evolution of stars. Ostriker, Charles A. Young Professor of Astronomy, is now using computational tools to understand the structure and origin of the universe.

"In many disciplines, we think we know what the equations are that govern the real world," Ostriker says. "But we don't know the solutions to those equations." In cosmology, for instance, scientists don't know what the initial conditions were when the universe was formed. Computer models allow cosmologists to plug different ideas about those initial conditions into a computer, then see if they generate a picture of the universe that matches current observations. "It's an essential tool," Ostriker says.

Immune system simulation

In molecular biology, scientists have encountered similar complexities in understanding the structure of proteins, the mechanisms by which drugs work, and how people and animals mount immune responses to invading organisms.

For example, biologists have reached the point where they understand the rules that govern individual components of the immune system but can't predict how they all interact without the use of computers, says professor of molecular biology Martin Weigert. Steven Kleinstein, one of Singh's PhD students, works with Weigert to develop a computer simulation of the immune system, which already has yielded insights and proven to be a valuable teaching tool.

Now Kleinstein, who spends much of his time with molecular biologists, is facing some of the dilemmas Singh experienced as a graduate student, but Singh reassures him about the value of interdisciplinary work. That message is also being sent on a departmental level. The University recently hired a new faculty member, Mona Singh (no relation to Jaswinder), who specializes in computational molecular biology.

PICASSO, which stands for Program in Integrative Computer and Application Sciences, will recruit PhD students from computer science as well as from other departments and involve the participation of local national labs such as the Geophysical Fluid Dynamics Labs and the Plasma Physics Lab, as well as industrial research labs. In addition to integrative training, the students will receive all the conventional training in their fields. They will receive certificates in Integrated Computer and Application Sciences along with their degrees.

PICASSO is funded by a five-year, $2.7 million grant from the National Science Foundation. The grant is part of the foundation's Integrative Graduate Education and Research Training (IGERT) grant program, designed to foster interdisciplinary research training. Princeton was among 21 institutions that received IGERT grants this year.