Professors Place Hoods on Newly Minted Ph.D. Graduates
|Schismenos, Ahipasaoglu, Leventhal, and Das following the Hooding Ceremony.|
photos: University Photography
Selin Damla Ahipasaoglu, Bikramjit Das, Dennis Leventhal and Spyridon Schismenos will officially receive their Ph.D.degree certificates in August, but they have already been provided with an important symbol of their accomplishments. At a graduation ceremony in Rockefeller Hall's Schwartz Auditorium on May 23, 2009, hoods were placed on the four, symbolizing their new membership in the doctoral ranks.
At the ceremony, Professor Michael Todd placed a hood on the shoulders of Damla Ahipasaoglu in his capacity as her Ph.D. advisor. Her dissertation is titled "How to Solve Ellipsoidal Inclusion and Optimal Experimental Design Problems? Theory and Algorithms." Ahipasaoglu was born in Erzureum, Turkey and grew up in Ankara. She came to Cornell from Bilkent University in Ankara and has joined Princeton University as a post doctoral fellow. Ahipasaoglu said that "Cornell was a great place to study, full of challenging and rewarding opportunities and remarkable people. I am very proud to be a member of such a special community. Ithaca will always be my home."
|Professor Michael Todd places Dr. Damla Ahipasaoglu's hood.|
Ahipasaoglu worked on a geometric problem with applications to data analysis, statistics, computational geometry, imaging and optimization. She improved the properties of an algorithm that computes the ellipse of minimum volume that encloses a given set of data points -- but can do so for an arbitrarily large number of dimensions (where the n-dimensional generalization of an ellipse is an ellipsoid). Such problems arise in machine learning, detection of statistical outliers, cluster analysis, and optimization. Her algorithm also computes solutions to an important and equivalent statistical estimation problem. She has shown that her algorithm runs quickly and accurately on very large problems, and has solved the largest known problem in the literature (10,000 data points in 500 dimensions) in less than 30 minutes.
|Dr. Schismenos prepares to receive his hood from Professor Henderson.|
Professor Shane Henderson hooded Spyridon (Spyros) Schismenos, having co-advised advised his thesis work, "Using Low-rank Matrices to Approximate Optimization Problems with Ellipsoidal Constraints: a Probabilistic Analysis," with Professor Adrian Lewis. Schismenos is from Panaitolio, a village in western Greece. A graduate of the National Technical University in Athens, he has accepted employment in the Credit Portfolio Group of JPMorgan, in London, UK. He commented that "ORIE was great and very friendly. I will definitely miss Ithaca, and hope to visit soon!"
Schismenos' work is motivated by a problem in designing radiation therapy plans, in which high doses must be administered to malignant tumors while minimizing the dosage to nearby organs despite the fact that the precise locations of targets and untargeted tissue can only be estimated and are subject to movement prior to treatment. A fully formulated model of this problem turned out to be computationally intractable, so the model was transformed into an approximation that was easier to solve but yielded results that were meaningful for the original problem. For his thesis, Schismenos created a stylized optimization problem similar to the radiation problem, investigated two simpler approximations to the problem he created and developed theoretical results about the relationship between solutions to the approximate problems and the original problem that help explain the success of the approach used in the radiation context.
|Dr. Bikramjit Das is hooded by Professor Sidney Resnick.|
Professor Sidney Resnick placed a hood on Bikramjit Das, his Ph.D. advisee. His thesis is on "Conditional Extreme Value Models." Das will be a post doctoral fellow at the Swiss Federal Institute of Technology in Zurich, known as ETH. Das is from Kolkata (formerly Calcutta), India, and came to Cornell from the Indian Statistical Institute in Kolkata. He said that "coming straight from India, I never felt an outsider here. The faculty in the ORIE department has been very kind, cordial and approachable." He has "realized through the last five years that very few places would have been able to provide the academic and social support that I received here. I will cherish my time in Ithaca."
In his thesis, Das extends the applicability of extreme value analysis, a statistical approach that is used for extrapolation well beyond the range of the available data. Extreme value analysis is used in many risk-sensitive fields such as environmental science (designing dams for a "100 year flood" and setting limits on atmospheric pollutants ); internet traffic data (which is subject to bursts, transmission of large files called "elephants", and phenomena that are correlated across multiple time scales); and financial risk analysis, where rating and governmental agencies require loss probability estimates at the 99th percentile (called "value at risk.") In estimating the probability of extreme values, analysts must rely on data sets in which no sufficiently extreme data points occur, Das focuses on extreme value problems requiring multidimensional observations, the components of which may not be independent of each other. For example an observation may describe the multivariate returns for a portfolio consisting of many securities, all dependent on the general economic climate. While standard mathematical approaches assume that all of the variables exhibit extreme behavior, Das analyzes the situation in which this assumption can be relaxed for some of the components.
Although Dennis Leventhal's Ph.D. advisor is Professor Adrian Lewis, Professor Todd (a member of his Ph.D. comittee) placed Leventhal's hood instead. At the time of the ceremony, Lewis was in the process of welcoming his new daughter, Nia Ylan Chi Lewis, at Cayuga Medical Center and so was unable to attend. Leventhal 's thesis title is "Effects of Conditioning on the Convergence of Randomized Optimization Algorithms." He is from Old Bridge, New Jersey, and holds a B.S. in Mathematical Sciences from Carnegie Mellon University in Pittsburgh. He will join Goldman Sachs in New York as an Associate.
|Dr. Dennis Leventhal and Professor Michael Todd.|
Leventhal's thesis deals with how quickly certain computational algorithms for mathematical optimization reach a solution. Such algorithms proceed step by step through a sequence of solution candidates. In the field of numerical analysis, speed of convergence is often related to how sensitive the problem is to perturbation of the input parameters, known as the 'conditioning' of the problem. Leventhal examines a class of algorithms in which the next computational step has an element of randomness that is determined by specific properties of the current solution candidate. He determines upper bounds on the rate of convergence of some randomized algorithms, with the objective, not of devising practical solution methods but of providing insight into the relationship between the conditioning of a problem and its computational tractability.