Archived post

This is an archived post. Please click here to see the latest entries.

« Beechert: Freshmen, Bring a Mask | Home | BREAKING: Hanlon Cancels Bonfire »


Brian Solomon’s Guide To The Stars: Engineering Professor George Cybenko

Dartmouth has a wealth of experienced professors who lead their respective research fields, while also working closely with students — inspiring them in the classroom and leading them in laboratory environments. And while at Dartblog we talk frequently about problems that need to be fixed at the College, there are still many bright spots. Our professors deserve more recognition for their achievements. As such, this is one of a series of posts that shines a spotlight on the best professors in Hanover:

George Cybenko.jpgGeorge Cybenko is the Dorothy and Walter Gramm Professor of Engineering at the Thayer School. His research has been at the forefront of influential mathematics and computer science fields like neural networks, signal processing, and distributed computing.

Cybenko grew up in Toronto and earned his B.A. in mathematics from the University of Toronto. Math always came easily to him, so he continued on to earn his Ph.D at Princeton. For his dissertation he worked on the digitalization of signal processing, the advanced method of filtering and transforming data so that it can be used by computers. As computing became widespread, analog circuitry in telephones and radar was replaced by digital approaches. Cybenko conducted pioneering research on how these signals could be processing. This work led to his interest in computers.

After graduating from Princeton in 1978, Cybenko joined the math department at Tufts, but brought along his love of computing — he would later serve as the founding chairman of the new Tufts computer science department. After ten years in Medford, he moved on to the University of Illinois at Urbana-Champaign, where he made another pivot into the electrical engineering school. The exact department in which Cybenko finds himself at any given time has never hemmed him in. Rather, as Cybenko says today, the boundaries between the fields of math, computer science, and electrical engineering are “fuzzy.”

Leaving Illinois, where Cybenko mostly did research and taught graduate students, coming to Dartmouth in 1992 was a welcome change. He downsized from a place with 100 electrical engineering professors to Thayer with 35 faculty in the whole school, but he was happy to be able to teach undergrads again (he calls them an “antidote” to cynicism).

Meanwhile, Cybenko produced an immense of high-quality research. He has 16,000 individual citations and an h-index of 40, according to Google Scholar. Some of his most cited works were written well before their time. His 1989 article, Approximation by superpositions of a sigmodial function, proved that simple neural networks could solve any computational problem. Interest in such computing systems modeled after brain activity died down after the early 1990s, but has recently had a large resurgence, with Facebok, Google, and Microsoft betting their futures on deep learning and A.I. advances based on neural networks. His theorems even have their own Wikipedia page. (Cybenko is thinking about going back to research in that field more.) Another 1989 paper, Dynamic load balancing for distributed memory multiprocessors, is relevant to swarm behaviors, where a flock of birds, a hive of bees, or, perhaps in the near future, a network of autonomous vehicles talk to one another.

Cybenko is one of several Thayer professors to help develop a pattern-recognizing technique called Process Query Systems that has multiple functions, but the work has led him specifically to investigate abnormal computing behaviors and hacking. Along with former Thayer professor Vincent Berk, Cybenko cofounded a company in 2004 called FlowTraq, which uses proprietary algorithms to analyze web network traffic patterns and detect problems. FlowTraq works with large companies in the defense industry and in online gaming to deal with massive amounts of data and commensurately large security concerns. Cynbenko has also created algorithms that learn to identify anomalies on their own, even when the computer doesn’t know at the outset what the pattern should look like. His experiments have shown that computers can analyze and identify suspicious trading patterns in NASDAQ data, even tracing the pattern to one particular trader.

Currently, Cybenko is in the middle of another five year project on computer security. The idea is that hackers have an advantage because if they find a single exploit in a popular program, they can leverage it to attack thousands or millions of sitting duck computers. Cybenko is researching ways to make computers moving targets by making every operating system different enough that hackers can’t attack them all. Such technology would need to be balanced with functionality — the greater the diversity in programming, the more difficult it would be for an IT department to manage its network of devices. Read more about Cybenko’s prolific history and work in this story in Dartmouth Engineer Magazine, and watch his lecture on challenges in cybersecurity here:

In addition to all his research, Cybenko typically teaches a mix of graduate and undergraduate courses, including ENGS 27: Discrete and Probabilistic Systems. He is the founding editor-in-chief of three publications: IEEE/AIP Computing in Science and Engineering, IEEE Security and Privacy, and IEEE Transactions on Computational Social Systems. In the past Cybenko has served on the Defense Science Board, on the US Air Force Scientific Advisory Board, and on DARPA review and advisory panels. He also has held visiting appointments at MIT, Stanford, and Leiden University.

Best-of-Logo-2014_dartblog.gif

Featured posts

  • August 14, 2013
    Breaking: Of Crips and Bloods and Memories of Ghetto Parties
    History repeats itself, first as tragedy and then as farce, or sometimes it just repeats itself. From the New York Times on November 30, 1998: At Dartmouth College, white students at a ”ghetto party” dressed…
  • June 25, 2013
    Dean of the College Charlotte Johnson’s War on Students Part (2/2)
    Part 1, Part 2 Today’s post again recounts the events that befell the Freshman. However, the content of the Hanover Police department report reproduced in this space yesterday is supplemented by information from my own…
  • October 18, 2009
    When Love Beckoned in 52nd Street
    We were at San Francisco’s BIX last evening, enjoying prosecco, cheese, and a bit of music. A full year of inhabitation in Northern California has unraveled to me no decent venue for proper lounging, but…
  • October 9, 2009
    D Afraid of a Little Competish
    So our colleague and Dartblog writer Joe Asch informed me that the D has rejected our cunning advertising campaign. Uh-oh. The Dartmouth is widely known as a breeding ground for instant New York Times successes,…
  • September 4, 2009
    How Regents Should Reign
    As Dartmouth alumni proceed through the legal hoops necessary to defuse a Board-packing plan—which put in unhappy desuetude an historic 1891 Agreement between alumni and the College guaranteeing a half-democratically-elected Board of Trustees—it strikes one…
  • August 29, 2009
    Election Reform Study Committee
    If you are an alum of the College on the Hill, you may have received a number of e-mails of late beseeching your input for a new arm of the College’s Alumni Control Apparatus called…

Dartblog Specials

Subscribe by Email

Enter your email address:

Help, Pecuniarily

Please note

This website reflects the personal opinions of its authors. Any e-mails received may be published along with the full name of the sender. If you wish otherwise, please say so.

All content appearing at Dartblog.com should be presumed copyright 2004-2017 its respective bylined author unless otherwise noted or unless linked to original source.

Advertisement



admin

Calendar

October 2017
Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30 31

Search

Archives

Links