LIDS all
Laboratory for Information and Decision Systems MIT
Big Ideas Big
Dreams & more
LIDS all
Editors:
Jennifer Donovan
Becky Karush
Irene Keliher
Design and Photography: Michael Lewy
Writers:
Meghan Dunn
Becky Karush
Alexandria Marzano-Lesnevich
Kate Willsky
Massachusetts Institute of Technology
Laboratory for Information and Decision Systems
77 Massachusetts Avenue, Room 32-D608
Cambridge, Massachusetts 02139
http://lids.mit.edu/
send questions or comments to
A Message from the Co-Director
Welcome to the Fall 2008 volume of LIDS|ALL. This issue provides windows through which we hope to give you a picture of the LIDS community – activities that make LIDS a vibrant and leading research organization, our research partners who are so crucial to extending and enriching our community, and the staff, students, and faculty who make LIDS an exciting and energizing place in which to work, learn, and create.
In this issue you’ll read about one of LIDS’ signature annual events, the LIDS Student Conference, an activity initiated and sustained by our graduate students and one that attracts many from the broader MIT community as well as distinguished speakers from around the world. You’ll also be introduced to one of our research partners, Draper Laboratory, through an interview with Dr. Eliezer Gai. Eli and many of the engineers and researchers at Draper Lab have been and remain very good friends and collaborators with many of us in LIDS, and
ABOUT LIDS
The Laboratory for Information and Decision Systems (LIDS) at MIT, established in 1940 as the Servomechanisms Laboratory, currently focuses on four main research areas: communication and networks, control and system theory, optimization, and statistical signal processing. These areas range from basic theoretical studies to a wide array of applications in the communication, computer, control, electronics, and aerospace industries. LIDS is truly an interdisciplinary lab, home to about 130 graduate students and post-doctoral associates from EECS, AeroAstro, and the School of Management. The intellectual culture at LIDS encourages students, postdocs, and faculty to both develop the conceptual structures of the above system areas and apply these structures to important engineering problems.
the support of Draper for LIDS and our activities has been of exceptional value to us.
You’ll also be introduced to some of our current and past students – Kush and Lav Varshney, both of whom are currently working toward their Ph.D. degrees, Lillian Dai, who received her Ph.D. this past June, and Sommer Gentry, who received her Ph.D. several years ago – and through them get a glimpse into the extraordinary set of students who reside in LIDS, make it a far better place, and then go on to distinguished and remarkable careers.
In addition, you’ll meet Lisa Gaumond and through her see that the LIDS administrative staff is itself not only a critical component of LIDS activities but also a group of creative and interesting people who add considerably to our environment. You’ll also encounter Henk Wymeersch, a post-doctoral researcher in LIDS, who, through that position, provides an invaluable resource to both students and faculty in terms of research leadership and execution. Finally, you’ll read about one of our faculty members, Pablo Parrilo, who is recognized as one of the world-leading researchers in optimization and who brings energy, vision, and technical excellence to LIDS’ table.
It would be wrong to say simply that LIDS is enriched because of activities and people such as these. LIDS exists because of activities and people like these. Thanks to them and the others in our community, LIDS stands as a world-leading center for research in information and decision systems and all of the academic disciplines that are crucial to this broad area of inquiry. The research at LIDS continues to provide not only seminal contributions in basic research crucial to problems of national and worldwide importance but also a training ground for the next generation of leading researchers and engineers whose contributions will undoubtedly make all of us at LIDS proud and provide great service to the world.
I hope that you enjoy this window into the LIDS community, and we invite you to learn more about us through our webpage or by visiting us. You are always welcome.
Sincerely, Alan Willsky, Co- Director
Draper Laboratory, Inc., and LIDS: A Win-Win Proposition
By Alexandria Marzano-Lesnevich
Draper Laboratory, Inc., is a familiar name to many at MIT. Draper’s name comes up in conversation in departments throughout MIT, owing to the wide breadth of the Laboratory’s respected Draper Fellows program and other initiatives in sponsoring research at MIT. The L ab’s black bulls eye logo is easily visible from the streets of Technology Square. But what is Draper’s relationship with MIT— and with LIDS? Recently, LIDS/All had the chance to talk with Dr. Eliezer Gai, Draper’s Vice President of Engineering, to learn more about this relationship.
Across MIT, Draper Laboratory plays a crucial part in providing students and faculty with the resources they need through the Draper Fellows program. Together with MIT faculty, Draper handpicks Fellows for funding and research support from the pool of admitted MIT graduate students. One such student is Sameera Ponda, a Draper Fellow who is currently finishing her M.S. thesis on mission planning for small unmanned aerial vehicles with Dr. Emilio Frazzoli, an Associate Professor of Aeronautics and Astronautics and member of LIDS. Ponda’s research is representative of Draper’s strongest connections to MIT. “The two main units Draper works with at MIT are Aero and LIDS,” says Dr. Gai.
It is neither an accident nor a surprise that Draper is so deeply entwined with MIT and with LIDS in particular. In fact, Draper Laboratories began life as part of MIT. In 1928, Dr. Charles Stark Draper became a professor of aeronautical engineering at MIT after completing his graduate work here. In addition to researching aircraft engineering, Dr. Draper was a civilian pilot, a hobby he acquired in order to gain insight into the practical applications of his research. It also made him increasingly interested in developing new instruments for navigation. Knowing the strength of the people who surrounded him at MIT, Dr. Draper decided to put his colleagues and students to work on these navigational tools and founded the MIT Instrumentation Lab in the 1930s. For forty years the Lab flourished. It provided the United States military with such cru-
cial innovations as the highly accurate inertial navigation system for ballistic missiles. It also conducted groundbreaking work for NASA, most notably, the Apollo program. The onboard guidance, navigation, and control systems it designed safely guided astronauts on the first trip to the moon in 1969. As tremendous as this progress was for the United States space program, the late 1960s also brought political unrest to the Lab. With the advent of the Vietnam War, the Lab’s focus on developing technology with military applications raised ideological dilemmas about the scope and purpose of uni-
the University. To continue the relationship, however, its employees would be able to use University facilities in exchange for an annual fee to be paid by the Lab. The Laboratory was so successful that it soon outgrew its MIT location. In 1975, Draper moved off campus into its own sprawling headquarters in Technology Square, where it stands today.
Because it was no longer housed on university grounds, however, the move prompted another dramatic change. Owing to the sensitive nature of much of the research conducted in its labs and the
The influx of new students and new professors means a constant infusion of fresh ideas to fuel Draper’s storied push toward innovation.
versity research among students and faculty alike; major concerns about possible use of scientific and technical knowledge in the context of the war effort were being brought to the fore. “There were daily demonstrations on the MIT campus,” Dr. Gai says.
“The students didn’t like the Instrumentation Lab to be involved with the war, and the faculty didn’t like approving dissertations [that might be associated with the controversial war effort].” Still, the research the Lab was conducting and the opportunities it afforded MIT students were too valuable simply to shut down. A compromise was struck: the Instrumentation Lab would cease to be a part of MIT, and would instead become an independent, not-for-profit research laboratory to be named after its founder, Dr. Draper. The new Laboratory would remain on MIT’s campus for the time being, but would no longer have any formal affiliation with
need for security clearances, Draper Research Fellows were now required to be United States citizens for Draper to be able to compete for government contracts. This meant that Draper could no longer support foreign students on its own contracts. But decades of affiliation with MIT had taught the Laboratory something—the strength of any institution lies in its people. Draper didn’t want to lose the ideas MIT’s foreign students had, and it didn’t want to leave the students at a disadvantage for funding to carry out their research. Draper decided to take $2 million from its own funds and give grants to MIT faculty members so they could sponsor their students’ research on campus without concern for nationality. It continues this generous funding today, giving away one percent of its revenue in grants each year. Such creative cooperation, Dr. Gai says, is just one example of how extensive the relationship be-
tween the two organizations is, and how much both sides benefit. Draper brings money in to MIT, he notes, a crucial concern for any academic institution. Its close proximity to the campus also indirectly expands the breadth of MIT’s research facilities, as the university doesn’t have to replicate the laboratories that are literally available just down the street. In terms of the actual research, Dr. Gai says, “Draper gives students and faculty the chance to do handson work, not just theoretical studies. The reason why Doc Draper actually founded the [Instrumentation] Lab to begin with was to give faculty and students the chance to do customer-relevant work.” Draper gets something important from the relationship, too. “We have access to the best students at MIT,” Gai says, “and therefore the best potential employees.” The influx of new students and new professors means a constant infusion of fresh ideas to fuel Draper’s storied push toward innovation. “Some of Draper’s biggest achievements are based on theses by MIT students,” Dr. Gai says, such as a “fault tolerant parallel processor” that’s been further developed for use by NASA’s Crew Launch Vehicle.
Such successes make it easy to understand why Dr. Gai is so excited about future collaborations between Draper and MIT—and LIDS in particular. “From a theoretical point of view, LIDS is probably the closest laboratory to what we are doing,” Dr. Gai says. An expanded relationship between Draper and LIDS will, Dr. Gai believes, make Draper even more competitive as it redoubles its efforts to win more government contracts. And more contracts for Draper means more opportunities for MIT students. “It’s a win-win situation. Everybody ben-
efits.” With this outcome in mind, Dr. Gai has been taking time out of his busy schedule to talk with LIDS faculty and students about research opportunities at Draper Laboratories. “Instead of [LIDS] just responding to Draper’s needs [for research] in an ad-hoc fashion, we will work year-round on that relationship,” he says. Dr. Gai was an invited speaker at the 2007 LIDS Student Conference, and more recently he and his Draper colleagues have been meeting with LIDS faculty at weekly lunches, where he explains the new thrust of Draper’s collaboration with LIDS. He hopes to align efforts between the two organizations, possibly structuring a joint team that would compete for contracts in areas as diverse as signal processing, networking, human-computer collaboration, and communications. “We need [LIDS],” Dr. Gai says. “And we feel they need us.”
Dr. Eli Gai, Vice President for
In the Swing of It: Adventures in Optimization
by Meghan Dunn
LIDS alumna Sommer Gentry remembers being scolded in her high school math class for explaining a problem to another student. She smiles, recalling: “the teacher said, ‘If you want to teach math, go out in the hallway and have your own class.’ And so we ended up having a little mutiny, and we sat in the hallway, and I actually did teach a math class.” She pauses. “I feel bad about it now,” she says. “It was probably a bad day in my math teacher’s life.” Maybe so, but it was a good day for Sommer Gentry, her future students, and her research. And it represents the talent and drive that eventually brought Sommer to LIDS and now to her teaching position at the US Naval Academy.
Sommer’s research has been both unique and rooted in real-world problems, albeit quite different -- swing dancing and kidney donation, for instance. And how do these connect? For Sommer, each holds fascinating research possibilities and dovetails nicely with her work in optimization, not to mention her passion for teaching and her determination to make a difference.
incompatibilities. Sommer’s research has an answer. In her scenario, this incompatible pair is matched with a second incompatible pair, ideally resulting in successful transplants for both patients. For example, if A is a willing but incompatible donor for B, and C is a willing but incompatible donor for D, A can give his kidney to D, and C can give hers to B. This way, donors who might not otherwise have
“I want to build something that changes the world,” she says. “I don’t want to write papers that sit on a shelf. I want to make a difference that I can see.”
Originally from Thousand Oaks, CA, Sommer earned her B.S. and M.S. at Stanford University before moving east to MIT in 1999. After finishing her Ph.D. work at LIDS in 2005, Sommer accepted a job as an assistant professor in the mathematics department at the US Naval Academy. She lives in Baltimore with her husband, swing dance partner and now research partner, Johns Hopkins surgeon Dorry Segev. Together, they work to incorporate mathematical tools into the kidney paired donation (KPD) system, hoping to increase the number of compatible donor-recipient matches in the United States.
All too often, a patient waiting for a kidney transplant may have a willing donor who is unsuitable because of blood type or tissue
volunteered to help a stranger have incentive: another stranger is helping their loved one. In other words, says Sommer, it’s “a win” for everyone. “We’re adding two donors to the net pool,” she says, “who wouldn’t have donated otherwise.”
Kidney paired donations like these have been performed at certain hospitals for several years, but Sommer points out, “there is no national system for paired donations.” According to the United Network for Organ Sharing, with whom Sommer and Dorry collaborate, there are currently about 70,000 people in the United States waiting for a kidney transplant. This makes Sommer’s research particularly pressing.
Here’s where Sommer’s optimization expertise enters the scenario. Currently, donor-recipient pairs are matched using a scheme called First-Accept Match, which locates and matches the first compatible pairs found. However, this scheme doesn’t take
into consideration all pertinent factors, such as other compatible pairs in the area and the ease or difficulty of matching them. Sommer explains that some transplant candidates are “highly sensitized,” or very difficult to match, and that by haphazardly matching pairs as in First-Accept Match, these candidates might be missing out on compatible donors.
To utilize all matching possibilities, Sommer advocates for a scheme called Optimized Match, which would consider all pairs in the national registry and make decisions “so that the most people could get a transplant, overall.” By using what is called “a maximum weighted matching,” the transplant community could decide “which pairs should exchange” in order to maximize the number of total pairs matched. Each donor-recipient pair would be represented by a node on a graph, and every compatible exchange would be designated by a line, or edge, connecting those two nodes. An optimized match contributes to the greatest possible number of compatible matches overall.
It’s an attractive, mathematical solution, but its practical implementation doesn’t come easy. Sommer explains that Optimized Match allows for caveats: “‘I’d prefer to do this match because this donation helps a child’ or ‘This donation helps a person who is very hard to match’ or ‘This donation is good because it’s an exchange between patients in the same
hospital,” and these considerations must be “translated into numbers.” This results in a complex pool of donors that surgeons and the transplant community must efficiently navigate.
To help bring Optimized Match to reality and attain her broader vision of “making optimization tools more accessible,” Sommer has reached further into her background. She’s using a tool called Inverse Optimization, the first topic she studied at LIDS. It would allow transplant surgeons to select their preferred matches from a list of candidates so mathematicians can convert those choices into “optimization parameters.” Sommer says, “if we do this several times, we’ll figure out what numbering scheme…will express their preferences” so these can be “codified into a system.”
It’s an exciting prospect. In an April 2005 paper published in the Journal of the American Medical Association, Sommer, Dorry, and co-authors concluded that a national organ sharing registry utilizing Optimized Match would yield “more transplants… than an extension of the currently used firstaccept scheme to a national level” and that “even if only 7% of patients awaiting kidney transplantation participated…the health care system could save as much as $750 million.” These kinds of numbers drive Sommer’s research and bring it positive attention. Three
important governmental rulings on KPD’s legality took place during 2007, bringing Sommer’s vision of a national registry closer to fruition. “I want to build something that changes the world,” she says. “I don’t want to
An unusual topic for LIDS, Sommer’s work traversed the fields of communications and optimization. As she became simultaneously immersed in the world of swing dance and the world of LIDS, Sommer began to think of the
“It’s
amazing how similar teaching dance is to teaching math”
write papers that sit on a shelf. I want to make a difference that I can see.”
In addition to this research and her teaching duties at the Naval Academy, Sommer somehow finds time for another passion: swing dancing. She and her husband Dorry (whom she met through dancing) teach the sport three nights a week with Charm City Swing, a Baltimore group they founded. As professional dancers, Sommer and Dorry have been honored nationally and internationally for their swing skills and take special pleasure in introducing non-dancers to the form. “It’s amazing how similar teaching dance is to teaching math,” she says. “In both cases, you have some students with a lot of anxiety about the subject.”
The connection between dance and math was the chief focus of Sommer’s Ph.D. research at LIDS, which looked at “the interactions between two people in a dancing couple.”
hand contact between two dance partners as haptic signals, messages sent by touch. “Our sense of touch is not all that well-developed,” she says, “so it’s strange that people could execute such elaborate movements…with such a bad communications channel.” Intrigued, and encouraged by her LIDS advisor Eric Feron, Sommer decided to explore further. “LIDS professors trust students to follow their own instincts,” Sommer says, citing this freedom as her favorite part of the LIDS experience, which helped her “grow to be an independent researcher.” She misses that environment along with the Stata Center, laughing as she recounts her struggle to “lay out furniture in an elliptical office.”
For the moment, Sommer plans to stay put in her (non-elliptical) office at the US Naval Academy. “It’s a good, supportive department and my students are great,” she says. “I am very, very happy here.”
PUTTING LIDS ON THE MAP:
POSTDOC RESEARCH INTO LOCALIZATION
By Meghan Dunn
When Henk Wymeersch arrived at LIDS from the Belgian university not far from his tiny hometown, he had a list of potential research topics “pages and pages long.” Rather than immediately focusing on one topic, Henk took the time to brainstorm with his advisor, Professor Moe Win, and decided to write a book on iterative processing, which he describes as a kind of “Iterative Processing for Dummies.” Iterative Receiver Design was published by Cambridge University Press in 2006. Henk, wiry and cheerful, is modest about this rather significant achievement. “It’s very exciting,” he says. “It was really nice of Professor Win to give me the freedom to pursue this.”
And that’s only the beginning of his LIDS experience, which has brought him far from Vinderhoute, Belgium, population 600. Henk first completed his B.S., M.S., and P.h.D. at the nearby University of Ghent. “Belgians don’t like to move very far,” Henk says, laughing. “Most people do undergrad, grad school, postdoc at the nearest university, and some eventually become professors there.” Henk broke with this tradition in 2005 when he was selected for a fellowship through the
that multi-pronged research centers on the problem of localization, which builds off Henk’s background in iterative processing.
Localization is what it sounds like: attempting to fix a location to someone, or something, in relation to other people or objects. When this person or object is mobile, localization is particularly tricky. Researchers must gather information about other objects or people in the area through an iterative, or repeated,
“I love working at a university. It’s impossible not to get involved.”
Belgian American Educational Foundation, which allowed him to continue his research at any U.S. university. He chose MIT, but says, “it was a nervous experience. I didn’t know what MIT was like. In my head, there were older buildings, brick walls, professors in old, book-lined offices. I had this picture in my mind, but every bit of it was wrong.”
Instead of those musty offices, Henk works in the Stata Center with Professor Win’s W Group (for wireless and wideband). The group’s interdisciplinary approach has allowed him to branch off from his Master’s and Ph.D. research and explore new areas. “Our group is special,” he says, “because we do theory, we do algorithms, and we do experimentation.” Right now, some of
process. Henk explains that an iterative process is a way of addressing a larger problem by breaking it into smaller pieces and using the solution to one piece to solve the next subproblem.
In the communications context, iterative processing is an efficient way of decoding bits of information that are sent from a transmitter to a receiver. Essentially, the receiver uses the pattern established by the bits of information it gets first to decode the bits that come later, allowing it to decode the entire message more quickly and accurately. Localization requires that the iterative approach go one step further, by moving the process from a receiver box to a network.
While localization has many potential applications, consumers are probably most familiar with its use in cellular phone networks, which allows participating users to see where their friends are. However, the current consumer technology for localization of this type still depends on communication between the mobile device and fixed base stations. If a cell phone is in range of three base stations, its location can be determined. Henk explains, “these base stations have known locations, since they don’t move. A cell phone can localize itself by measuring the distance between itself and at least three base stations.” But this existing architecture is ultimately limiting. If the device is out of base station range, localization isn’t possible.
Part of Henk’s current research addresses this problem. “I’m working on an architecture, which depends on both base stations and the network,” he explains. In other words, if one user in the network were able to localize by connecting to a base station, other users in the network, who might be out of range, could determine their own locations by connecting to the localized user. He elaborates, “Maybe you know where you are, but I don’t know where I am. I could talk to you directly, figure out my distance, my position with respect to your position. I don’t need to be able to talk to all three base stations.”
now, although he makes time for running on the beach in Dorchester and attending events sponsored by MIT’s EuroClub. He responds to his friends’ and family’s questions about his LIDS life by explaining, “I work seven days a week, basically. But this is the most amazing intellectual environment. I go to work every day with a big smile on my face. I feel very lucky.”
In the future, Henk hopes to keep going to work smiling. His goal in the next few years is to become an assistant professor at a U.S. university. “I love working at a university. It’s impossible not to get involved.” His advice to future postdocs? “A postdoc is what you make of it… grab as many experiences as you can.”
This research is Henk’s real passion right
2008 LIDS Student Conference
SPEAKERS
Emmanuel Abbe
Amir Ali Ahmadi
Professor Emery Brown
Amit Bhatia
Dr. Eliezer Gai
Emanuele Garone
Dr. Bill Irving
Sheng Jing
Sertac Karaman
Mr. Zachary Lemnios
Ilan Lobel
Nicholas Matsakis
Baris Nakiboglu
Roberto Naldi
Jelani Nelson
Urs Niesen
Paul Njoroge
Marco Pavone
Raluca Ada Popa
Victor Preciado
Robert Panish
Alexander Olshevsky
Mike Rinehart
Melanie Rudoy
Navid Sabbaghi
Dr. Richard Sears
Proffesor Jeff Shamma
Jay-Kumar Sundararajan
Jinwoo Shin
Noah Stein
Vincent Tan
Professor Don Towsley
Sankara Narayanan Vaideeswaran
Lav Varshney
Professor Alan Willsky
Theophane Weber
Bright Lights, Big Ideas
By Meghan Dunn
Some LIDS students have always known they were MIT-bound. For Lillian Dai, however, the road to LIDS has been a long and winding one. From China to Canada and now Cambridge, Lillian’s love for learning has brought her to diverse fields of study, a wide range of extracurricular interests, and a belief in the interdisciplinary nature of LIDS.
Lillian was born in Xining, in the Qinghai province of China near Tibet, an area she calls “remote,” and moved to Calgary, Alberta at age eleven with her family. The transition was difficult for her. “I didn’t speak a word of English,” she says. “I knew my ABCs, but that was it.” Though her English skills were limited, Lillian excelled in math, a subject that needed no translation. As Lillian’s English improved, she also shone in other subjects and briefly considered careers in medicine, environmental engineering, and music before enrolling at the University of Calgary to study electrical engineering. During her fourth year there, Lillian worked with a professor on a satellite communications project, and though she had little experience with communications, she quickly became hooked. When she learned that members of the LIDS faculty were researching satellite communications, she applied. Before making her decision, Lillian came to visit the lab and to meet with faculty. “I was so impressed,” she says, “I decided right then that I was coming to LIDS.”
Lillian’s research emphasizes an interdisciplinary approach. “I think it’s the right thing to do,” Lillian says. “When you design a communications system, you can’t just focus narrowly on one small aspect; you need to look at the whole picture.” That picture may include disciplines beyond communications, as in Lillian’s master’s thesis, which addressed satellite link design from a cost perspective. Her current PhD research requires an even greater perspective. Lillian works to resolve the “inherent problem” of unreliability in existing infrastructureless wireless networks. She describes the scope of the problem as “huge,” involving not only communications and networking, but also optimization, control, and signal processing. In short, “everything at LIDS.”
Infrastructureless wireless networks, like those Lillian researches, are temporary wireless networks that can be deployed rapidly anywhere and anytime “without having to deploy fixed network infrastructures ahead of time.” This means that no fixed base stations or access points need to be built and installed, unlike cellular and Wi-Fi networks. Like walkie-talkies, the devices in an adhoc network communicate via radio waves and can only “talk” to other devices within a particular distance. Unlike walkie-talkies, each device “acts as a relay for other people’s communication” in order to extend the range of the network. Information “hops” from one
mobile device to the next, until it reaches its intended destination; the more devices participating in the network, the greater the potential for extending the network’s range. As people move, along with the devices that they use, the topology, or geographic structure, of the network continuously adjusts, changing the route that information will take.
One type of infrastructureless wireless network under intensive study is the adhoc wireless network, which, as the name implies, offers ad-hoc or best-effort service to devices in a network. As devices move, ad-hoc
disaster relief coordination issues had to do with a lack of communication.” By reliable infrastructureless wireless networks, she suggests, rescue workers and police could have maintained communication throughout the devastated city.
Creativity and interdisciplinary thinking provide potential solutions to the problem of unreliability in existing ad-hoc wireless networks. Lillian proposes “a new infrastructureless wireless network architecture called Proactive Mobile Wireless Networks,” an approach that goes
“When you design a communications system, you can’t just focus narrowly [on] one small aspect; you need to look at the whole picture.”
wireless networks may become disconnected, and hence cannot guarantee that time-critical information will reach the intended recipients in a timely manner. For many applications such as communication during disaster relief efforts, search and rescue missions, and special operations in the military where timely delivery of information is often a matter of life and death, new wireless architectures must be developed. Lillian names the aftermath of Hurricane Katrina as a key example. “All of the communication infrastructures were down,” Lillian points out, and “a lot of the
beyond improving the usual communication resources by incorporating proactive network state prediction and adaptive helper node insertion and trajectory control. These helper nodes could be fixed relay devices randomly scattered in a region or mobile access points mounted on vehicles, hot air balloons, or in the future, even robots. With the ability to predict disconnection times and locations, and with helper nodes enabling connection among devices that would otherwise be disconnected, there is a better chance that all the participants in the network can stay connected and
messages communicated successfully. Yet Lillian’s problem doesn’t end here. The other half of her research involves optimizing communication and network resources at each device, by using more advanced antennas on mobile devices and finding energy efficient routing paths to prolong battery life-time.
“I have built the architectural framework,”
Lillian says of her research. “Eventually it will take a lot more effort than just me alone to refine the details.”
Lillian’s passion for knowledge doesn’t end at the lab. “I really feel that to be a good engineer you need to be well-rounded,” she says of her many extracurricular activities.
Lillian’s advice to future students echoes this sentiment: “Be curious: MIT is like a treasure box and you should pick up as many treasures as you can.” In the past few years, Lillian has
been an officer in the Science and Engineering Business Club, served as music chair and trustee for the Sidney-Pacific dormitory, and founded both MIT VentureShips, an organization that connects students seeking business experience with MIT-affiliated start-up companies, and MentorConnection, a one-to-one mentoring program which matches students with successful MIT alumni.
Lillian’s newest project is the creation of a LIDS Facebook group, in order to increase interaction within the lab. Using Facebook, a social networking site geared toward students, LIDS students and staff can organize movie nights and dinners, or notify group members of upcoming LIDS soccer games. Lillian hopes more alumni will join the group as well, further enhancing networking opportunities.
Lillian also makes a point to attend different events and talks on campus. “That’s what makes MIT special,” she explains. “You can invite these great people and they will come and share their stories….When I first came here, I didn’t know very much about research or about the world. I felt like I was living in a fog and there were only bits of light coming through.” Her ambition and zest for knowledge guarantee those bits of light will continue to grow, and she hopes, brighten others’ lives. Lillian looks forward to collaborating with other talented young researchers in the future because, she affirms, “together, we can make a stronger impact.”
2007 Events @ LIDS
Maybe the desire to be an engineer runs in the blood. Maybe, like the genomes that Kush Varshney studies with graphical models, it clings to DNA and lures in proteins, becoming a transcription binding site and securing its continued existence in future generations. The Varshney family pedigree offers a prodigious case example of this theory.
Kush and Lav Varshney:
The Big Dreams of Twin Engineers
By Kate Willsky
“Our grandfathers and our father were electrical engineering professors,” says Lav Varshney. He and his twin Kush are both LIDS students working on their PhD theses in engineering. This legacy actually goes back more than just two generations. Years after fully launching themselves into the world of engineering, Kush and Lav discovered, through vague family whispers and some help from the MIT archives, that their great-great grandfather, Ishwar Das Varshnei, studied chemical engineering at MIT in the early 20th century. He went on to more or less found the glass industry in India.
Whether passed through blood or chance, the curiosity, drive, and innovation that produce a successful engineer have coursed through these past hundred years, running strong as ever to and through Kush and Lav. While Kush was more naturally interested in math and physics, and Lav admits that he was “actually better at social studies,” both chose to pursue engineering degrees at Cornell University. The twins graduated in the spring of 2004 and suspended their studies for a few brief months of travel, including some time in India and at the Olympics in Athens. Then they were Boston bound, arriving at MIT by summer’s end, ready to dive into their graduate studies.
“I think for me the faculty was the main reason I came here,” says Lav. In Central New York, Kush says, people “don’t necessarily have big expectations before doing something but are happy when it turns out well,” while at MIT, people
“are aggressive and have expectations of succeeding. They’re more goal-oriented.” This attitude is part of Boston itself, Kush muses. It’s a city where people “feel like they want to make it a big city,” and, by virtue of their desire, it becomes one.
In such an environment, big dreams and grand goals flourish, and pursuing them becomes not only possible but absolutely necessary. The variety of Kush and Lav’s projects over the past few years mirrors this drive. They have studied information theory and signal processing, radar imaging, graphical models, and data compression, and are working toward practical applications as diverse as finding trapped oil underground and mathematically modeling racism in the NBA.
For his Master’s thesis, Kush studied radar imaging, using synthetic aperture radar to develop technology that would allow airplanes to identify the irregular, distorted, light-reflecting objects far below.
Since then, “I’ve really shifted gears,” Kush says. His current work involves looking at graphical models to represent probability distributions. Studying random variables with some unidentified correlation, Kush attempts the formidable, somewhat paradoxical task of “figuring out the randomness.”
In a valiant effort to give form to this advanced, abstract mathematical concept, Kush supplies a plethora of examples, starting with the weather in
Boston and Syracuse. “They’re probably correlated in some way, like the temperature,” he says. Knowing the temperature in Boston can give you a pretty decent estimate of the temperature in Syracuse—but then add in India, a location with a completely independent temperature variable. A graphical model of this situation, with each place a different node in a graph with the nodes for Syracuse and Boston connected to each other but unconnected to the node for India, allows one to tease out correlations and understand the randomness.
“Actually, it’s a little more complicated,” Kush concedes, and offers DNA in cells and a videotape of a football game as other examples. He also mentions the work of the Stochastic Systems Group, which with recent funding from Shell Oil uses seismic data to create graphical models that reveal underground pockets of trapped oil. What these cases all emphasize, it seems, is the importance of recognizing patterns within a larger whole. “It’s about using what you’ve learned to identify what you’re given,” says Kush.
Lav, for his Master’s thesis, looked at the human
brain as “an information storage device” in an attempt to study the functioning of the human mind by maximizing information storage capacity, while also conserving resources, a challenge with potential applications for computers and databases (as well as human beings). Since completing his Master’s, he has been “playing around with various problems and they all seem to be related to graphs.” He’s interested in connections within networks, an idea that could apply to the Internet, genomes, and neurons in the brain, among much else. An example of the type of graphical connection he might find, Lav explains, is to take an image, segment it into pieces, and look at the way these pieces are attached.
Whether passed through blood or chance, the curiosity, drive, and innovation that produce a successful engineer are coursing strong as ever to and through Kush and Lav.
Ultimately, he says, “the question I’ve been looking at here is whether there are source coding theorems for graphs.” This line of research looks to compress data, squeezing all of the information from an ensemble of graphs into a simplified code, one requiring fewer binary digits. Using probability and general knowledge, one is led, eventually, to universal compression theory, the idea behind zip files, which states that one need not know the information in something to com-
press that information. “Zip is actually considered one of the great successes of an application of information theory,” Kush interjects, underscoring the importance of this project.
Still, these forays into research are tentative PhD ideas, and neither Kush nor Lav has definitively chosen a topic. “The deadline’s pretty flexible,” Lav says. But, as he wisely adds, the earlier you get started, the earlier you can finish.
Meanwhile, LIDS provides an active and stimulating intellectual environment in which to develop these ideas and discover new ones. Indeed, one of the most valuable aspects of the lab has been its ability to introduce students to the overall research process, Kush says, allowing them to start off with a “small idea…and build upon it and see where it goes.”
T he lab has also allowed both brothers to attend myriad conferences over the past few years, traveling to Florida, Utah, San Diego, and Lake Tahoe and hearing about current developments in the fields of biomedical imaging, data compression, and information theory. Lav calls these conferences “really, really tiring,” but emphasizes their ability to expose researchers to problems and questions they would not otherwise encounter. “You do learn a lot,” he says. “There are so many new ideas, and you get to talk to people and meet people.”
This commitment to bringing people and their
ideas together applies to daily life, too, as with LIDS’ Student Socials, a practice that Kush considers one of the great aspects of the lab. “It’s a good way to just get everyone out,” he says.
Socials aside, the brothers both lead active and diverse lives beyond the confines of the laboratory. Traveling scores high as a recreational interest, with Geneva, Berlin, Brussels, and Paris receiving honorable mentions as some of the most
As a newer discipline, electrical engineering seems, to Kush, “where all the action is.” It is a field in flux, the “liberal arts of engineering,” according to Lav. This latter quality directly appeals to Lav, who emphasizes the importance of applying math and science to human life. “You can do straight up information theory if you want to. The question then is, is it relevant to anything?” he asks, declaring that his engineering must be motivated by a concrete goal, however
LIDS provides an active and stimulating intellectual environment for the Varshneys to develop their ideas and discover new ones.
memorable and enjoyable cities for the brothers. While in Boston, (although there’s not a whole lot of time to relax, Lav admits), the brothers explore the city, attending sports games or social events with other graduate students. Relaxation, of course, also includes the simple joy of “just sitting back and watching TV,” says Kush.
Still, these respites are only momentary breaks in the action, commas within the constantly lengthening, infinitely complex, subordinate-clause-ridden run-on sentence that is electrical engineering. After all, it was the energy and dynamism of the field that attracted Kush and Lav in the first place.
far-removed or indirect it might be. Electrical engineering, by virtue of its abstract nature, can be applied to many things, allowing a freedom within the field, particularly an ability to look at human-scale implications that could lead to social or policy changes.
The experimentation allowed in electrical engineering seems analogous to the freedom and lack of responsibility of a university student. Kush and Lav recognize the unique privilege afforded a student and relish the uninhibited intellectual environment and relatively unstructured lifestyle.
“You don’t have to have a real job,” Lav says, smiling. “And we don’t want to grow up anyway!” But still, the pull of personal growth is as strong
as that of technological progress. Kush notes that, despite all the upsides, he would not want to stay a student interminably. “You feel like you should be graduating at some point, and doing some other things,” he says. Given what Kush and Lav have done so far, and the long line of innovation and talent that preceded them, we can only imagine what these “other things”—the fresh ideas and exceptional projects—will be.
The Pleasure of Finding Things Out
By Meghan Dunn
It may have been the controversial “new math” that sparked young Pablo Parrilo’s interest in mathematics. In the 1960s and 70s, educators in Argentina, where Parrilo grew up, supported the idea of emphasizing structural principles rather than rote memorization in elementary math classes. Since then, educators throughout the world have moved away from the “new math” concept, but Parrilo was addicted from a young age. The theories made it seem more appealing than traditional arithmetic. “I do remember that it was fun,” he says.
Today, Parrilo is a Professor in the Department of Electrical Engineering and Computer Science, is on the faculty at LIDS, and is affiliated with the Operations Research Center. He came to MIT four years ago, after working as an assistant professor of Analysis and Control Systems at the Automatic Control Laboratory of the Swiss Federal Institute of Technology in Zurich. He received his Ph.D. in Control and Dynamical Systems (CDS) from the California Institute of Technology in 2000, and though he lived in Argentina for most of his life before then, has experienced very little trouble adjusting to the American way of life. “The U.S. is an easy place to get adapted to,” he says. “This
country is well-prepared to receive an influx of people from different places.”
Pablo’s research at LIDS comprises mainly two areas: optimization techniques for engineering applications, and control and identification of complex physical systems. To put it more broadly, he studies systems in the real world, works to come up with mathematical models of them, and then he and other engineers can use these models to solve relevant engineering problems. Mathematical models are essentially formal descriptions of a system, and much of a researcher’s focus may be spent devising and improving these models, as well as designing algorithms for coming up with better representations.
While highly theoretical to most of us, for researchers like Pablo these models provide a tangible base from which to modify systems. These systems can be very concrete, such as a car engine, or more complex and less-defined, such as an economic or social network. “There are many cases where you have explicit mathematical models of a system,” Pablo explains, “and these models tell you, for example, how your car should run for optimal fuel consumption, or how to best stop the propagation of a disease.” However, he points out, describing a system using mathematics is not a simple task. While eminently useful for designers and system controllers, “these models are never exactly correct,” he notes. So, the question for researchers becomes, “how can you make accurate and reliable predictions, even though your models are only an approximation
to reality?” Expanding on the car example, Parrilo points out that there are many events that could happen while driving a car, and it is difficult to take into account all of these through simulations. Imagine the possibilities: a sudden stop, airbag deployment, or a dramatic change in terrain, for example. However, with increasingly accurate optimization techniques and mathematical models, researchers and designers can get closer to accounting for all such eventualities, hopefully leading to safer, cheaper, more efficient systems.
Parrilo deeply appreciates what he calls (in Richard Feynman’s words) “the pleasure of finding things out.” While not unusual at LIDS, he points out that such passion is not prevalent everywhere. Indeed, he’s concerned about the younger generation’s lack of interest in research, something he cites when asked about American students’ declining math scores. “I think the notion that knowledge should not be pursued for its own value is a terribly destructive notion. We need to know things beyond today or tomorrow.”
He also suspects many students decide too early to abandon a promising research career path in favor of earning more money. At MIT, however, he’s able to stimulate intellectual curiosity through his classes, which he enjoys a great deal. During the Spring 2008 semester, he taught “Algebraic Optimization and Semidefinite Programming,” a new course on advanced mathematical programming techniques. Parrilo says, “one of the things you always get out of teaching is a direct sense of accomplishment…by the end of the lecture the students, and you, have hopefully learned something.”
If he hadn’t become a professor and engineer, Parrilo says he may have followed his inclinations to have a career in the movies – behind the camera, as a director. He goes back to Argentina about once a year, where his sister is a veterinarian in Buenos Aires. Parrilo occasionally watches Argentinean football (soccer) during the World Cup when it comes around every four years, but is not a huge fan. “Some people claim that’s why they kicked me out of Argentina,” he laughs. Whatever the case, LIDS is very happy to have him. He is an accomplished, dedicated teacher and researcher, and has been recognized for that work in recent years. In 2005 he won the Donald P. Eckman Award from the American Automatic Control Council, and he has been a Finalist for the Tucker Prize of the Mathematical Programming Society for the years 2000-2003.
Sound Bites: Lisa Gaumond
By Becky Karush
What do you do at LIDS?
I’m an Administrative Assistant. Basically I’m part of the support staff, so I have four different faculty members I support. They’re a good bunch. A little feeding, a little watering, and they’re fine.
How long have you worked at MIT?
I started at LIDS in 2004 and at that time I was only supporting one person, but he soon left the institution. So I went to a temp job in the math department for nine months as a senior administrator, and then I went CSAIL. I also started a PhD program then, so I was only working part-time.
Then I came back here about a year ago. It’s interesting here because I’m exposed to things that I would otherwise really run away from.
Like what?
The math stuff, and the engineering things. The students will show me this amazing thing they’ve discovered and I’ll say, Wow, that’s pretty, and poke it. It’s a flux capacitator, they say. And I say, Ohhh. It looks like a paperweight.
I like the students a lot, even though I really have no idea what they’re talking about most of the time. You know, they are so modest! They’ll be doing something on a white board, some complicated pictures of numbers, and I’ll walk by and go, Oh god, no, wait, that’s—you’ll want to check that. And they’ll start checking it! And I’m like, Guys, it’s me, the social scientist.
What do you like about working in LIDS?
I think the smaller labs suit me. But I honestly didn’t even mean to take a real job at LIDS, once my first temp job assignment was up. I actually kind of tried to blow the interview because I knew that I liked it here and that if I stayed I would never go to a PhD program, so I showed up late to the interview, which was just wrong. And when the hiring professor asked if there was anything else he could do to make me more comfortable, I said, “How about a wet bar?” And then I found out that he happened to be a person in the department I could say that to who wouldn’t be offended. He
said something like, I was thinking we could do that in the electrical closet!
What do you like about being at MIT?
I love the media lab here. It really overlaps with a lot of work I’m doing, so much that it influences my PhD work. One of the competing topics for my dissertation is actually based right on technology here at MIT.
What is the PhD?
It’s a weird program, actually. I did my Master’s in forensic psychology, and I couldn’t decide whether to do research or clinical work. Then I found this program at a school called Fielding Graduate University with a PhD program in Media Psychology. It’s the only accredited program of its kind in the world that we know of!
You started in forensic psychology?
Well, I ended up officially doing criminal justice because forensic psychology wasn’t a Master’s option where I was. I got to visit crime scenes and see cool stuff at the crime lab in Boston, and hang out with the homicide unit. And once I even got to say, There’s nothing to see here, folks, move along. I really did!
What are your goals for the future?
To pay my mortgage? How far into the future are we talking? Seriously, I would like to— today anyway—write legal narratives for underrepresented populations, to be able to make really cogent arguments on their behalf,
or to make their characters come across.
Right now I’m also overly fond of Judge Judy. I absolutely love it! I like the idea of Judy having her agenda, which is to keep within the legal reasoning, but then you have people telling their narratives and the counter-narratives. Plus I like the intrigue of when somebody gets caught in lie. You can tell if the person feels guilt or is aware that, Ooh, I’ve been had.
So I like that it gets illustrated how people come to their own belief systems, and how they grasp or don’t grasp or refuse to grasp any other belief system.
And what about your black and white Boston Terrier, Iggie? Does he always come to work with you?
Oh, the man in my life? Yes, he’s here every day. It’s really interesting to see how all the dogs here get acculturated. On the train, it’s like a Monday for Iggie, too—he just snoozes the whole time, then he gets here and gets in the elevator with another dog, and they don’t look at each other, they just stare straight ahead. He used to have coffee every morning with someone in the math department. He has his own independent life. He keeps me on a need-to-know basis.
MILESTONES
Professor Alan Willsky celebrated his 60th birthday with a symposium held in his honor on May 31, 2008. Over eighty participants came together to recognize Professor Willsky’s contributions to the field as a researcher, entrepreneur, and mentor. Talks were held throughout the day on topics such as nonlinear filtering, signal representation, graphical models, and stochastic processes. The event culminated with an open microphone for personal reminiscences and comments from Professor Willsky.
LIDS is pleased to announce that Professor Sanjoy Mitter was the winner of the 2007 Bellman Award. This award is the highest recognition of professional achievement for US control systems engineers and scientists and is given for distinguished contributions to the theory or application of automatic control. (Professor Mitter is proudly continuing what is by now a tradition - former LIDS Director Mike Athans was a recipient of the award in 1995, and LIDS alum Roger Brockett won in 1989.) In his acceptance speech, Professor Mitter called receiving the award a “great honor” and spoke about the state of the field.
To see a full transcript of Professor Mitter’s speech please visit http://lids.mit.edu/lidsmag08/mitter
LIDS Staff Administrator Doris Inslee retired on July 31, 2008, after nine years of service to LIDS and over 25 years at MIT. Doris’s warm, vibrant, and caring presence will be greatly missed at LIDS - we wish her all the best as she embarks on her next adventures.
In July, 2007, Eileen Ng-Ghavidel, LIDS Administrative Officer, was recognized with a promotion to Assistant Dean for Finance and Personnel for the School of Engineering. She joined LIDS in 2001 and managed the lab with great success for six years, helping to strengthen the LIDS community and oversee times of transition. We wish her continued success in her new role.
LIDS Colloquia 2005-2006
Weekly colloquia are a highlight of the LIDS experience. Each talk, which features a visiting or internal invited speaker, provides the LIDS community an unparalleled opportunity to meet with and learn from scholars at the forefront of their fields.
The Stochastic Systems Group seminar schedule can be found at: http://ssg.mit.edu/cal/cal.shtml
Listed in order of appearance.
RESEARCH AREAS
Daniel Liberzon UIUC
Wei Yu University of Toronto
Michael Kearns (TOC) U. Penn
Terence Fine
Cornell
Mario Sznaier Northeastern
Erol Gelenbe Imperial College
Shuki Bruck Cal Tech
Vincent Poor Princeton
Kamal Jain (TOC) Microsoft Research
Giuseppe Caire USC
Lievan Vandenberghe UCLA
Mario Gerla UCLA
Calin Belta
Boston University
Kamesh Munagala Duke
Ben Van Roy Stanford Massimo Franceschetti UCSD
Bill Freeman MIT
Drew Fudenberg
Harvard
Eric Klavins U. Washington
Peter Glynn Stanford
Gustavo de Veciana U. Texas
Diana Dabby Olin
Jennie Si Arizona State
Imre Csiszar Renyi Inst.
LIDS News & Honors
Congratulations to our members for the following achievements!
Professor Mujdat Cetin , a LIDS Visiting Research Scientist and faculty member at Sabanci University, Istanbul, Turkey, received the 2008 Turkish Academy of Sciences Young Scientist Award.
Professor David Forney was awarded an honorary doctorate from Ecole Polytechnique Federale de Lausanne, Switzerland in 2007.
LIDS Staff Administrator Doris Inslee was a 2008 winner of the MIT Excellence Award for Serving the Client.
System Administrator Brian Jones was a recipient of the 2007 Steven Wade Neiterman Award.
Junmo Kim, Mujdat Cetin, and Alan S. Willsky received the 2007 Elsevier Signal Processing Journal Best Paper Award for their paper titled “Nonparametric Shape Priors for Active Contour-based Image Segmentation”.
Jaime Lien received the 2007 David Adler Memorial Thesis Award for Outstanding Electrical Engineering M.Eng. Thesis from the Department of Electrical Engineering and Computer Science, MIT.
Professor Asuman Ozdaglar was the 2008 winner of the Donald P. Eckman Award.
Professor Thomas Magnanti was the recipient of an honorary doctoral degree from Technion. Professor Magnanti was also awarded the Harold Lardner Prize for International Distinction in Operations Research from the Canadian Operations Research Society.
Professor Sanjoy Mitter won the 2007 Richard E. Bellman Control Heritage Award for contributions to the unification of communication and control, nonlinear filtering and its relationship to stochastic control; optimization; optimal control and infinite-dimensional systems theory.
Sue Patterson , Research Program Specialist, was a 2008 recipient of the School of Engineering’s Infinite Mile Award for Excellence.
LIDS Alum Sridevi V. Sarma , EECS Ph.D. Graduate (February, 2006), won two prestigious awards: the Burroughs Wellcome Fund Award and the 2008 L’Oréal USA Fellowships for Women in Science.
Professor Devavrat Shah was the recipient of the First ACM Sigmetrics /Performance Rising Star Award.
Yuan Shen and Henk Wymeersch received the Best Paper Award at the 2007 IEEE Wireless Communications and Networking Conference.
Noah D. Stein won the first place Ernst A. Guillemin Thesis Award for Outstanding Electrical Engineering S.M. Thesis, MIT.
Erik Sudderth , a recent LIDS grad, has been listed as one of the IEEE Intelligent Systems top “Ten to Watch”.
Professor John Tsitsiklis was a recipient of the INFORMS Fellow Award.
Professor Moe Win has been selected as an IEEE Distinguished Lecturer by the IEEE Communications Society.
http://lids.mit.edu/