ORIE Celebrates Professor Resnick’s 70th Birthday with a Roast and a Sidposium
Friends of Lee Teng-Hui Professor in Engineering Sidney Resnick were not surprised at the announcement that his birthday would be celebrated by a roast, which was held in the Herbert F. Johnson Museum of Art and attended by family, friends, and colleagues. Though for some attendees the event might have been considered “payback time” the evening was marked by warm affection tempered with (mostly) gentle gibes.
A month later, the celebration continued with a New York City Gala held at Cornell Financial Engineering Manhattan (CFEM). The Gala consisted of a series of technical talks by long-term Resnick colleagues, culminating in dinner (with more roasting) in a nearby restaurant dining room located in the antique vault of the former JP Morgan Bank.
The Ithaca Roast
Roasters included Professor Richard Davis from Columbia University; ORIE Professors Mark Lewis, Mike Todd, Shane Henderson, Franz Schalekamp and David Shmoys; a trio of members of ORIE’s entering Ph.D. class of 1993; and Resnick’s son Nathan ‘05, now a technology recruiter based in New York City.
The trio at left, former graduate student office-mates Victoria Averbukh Kulikov Ph.D. ’97, Kevin Wayne Ph.D. ’99, and Kathryn Caggiano Ph.D. ’98, reminisced about Resnick birthdays past that were marked by posters (some of which reappeared at the New York City Gala) and “anonymous” pranks such as filling his office with balloons. They presented Resnick with a new autographed poster featuring the cover of his 1992 graduate text Adventures in Stochastic Processes, a book enlivened by “rich vignettes of whimsically imagined Happy Harry and his Optima Street gang’s adventures in a world whose randomness is a never-ending source of both wonder and scientific insight.” The original cover shows Happy Harry running up a graph: the autographed poster shows Resnick drawing a bicyclist, probably himself, racing down the graph on a collision course with Harry.
Nathan Resnick (left) provided another high point of the roast, a talk in which he characterized his father’s unique sense of humor, “to clarify, I mean unique sense of humor for an adult. He is basically a Benjamin Button of humor – the older he gets the younger his sense of humor gets.” Nathan Resnick discussed some of the situational problems based on Happy Harry, noting that “I have no idea where my dad came up with them, but naturally, I always assumed that my dad modelled Happy Harry after himself and was into some super weird stuff before he met my mom [the artist Minna Resnick].” Nathan said that his father continues to come up with situational problems for his classes, some of which have been based on family situations, such as one about “Nasty Nate, a bartender at Rulloff’s, who leaves town and leaves Sid and Minna with a lot of beer that may or may not have gone bad, and now they don’t know which beer to serve at a grad student reception. (Although Nathan “Nate” Resnick actually was a bartender at Rulloff’s, a Collegetown restaurant when he was a student, he disavowed any connection to the “Nasty Nate” character in the 1998 film “Half Baked”).
Noting that both Sid and Minna Resnick are bicyclists and that Sid is especially fond of his small folding “Bike Friday,” ORIE Director David Shmoys presented the each of them with a small, highly detailed bicycle, with a suggestion about how they might be able to make them grow.
The New York City Sidposium
Most of the speakers at the New York City Gala have had a long professional and personal association with Professor Resnick. Thomas Mikosch (University of Copenhagen), Paul Embrechts (ETH) and Claudia Klüppelberg (Munich University of Technology) were at ETH, the Swiss Federal Institute of Technology in Zurich, Switzerland, in 1992 when Resnick met them as a Visiting Fellow there. These three speakers have written papers with him, as have the other speakers: Holger Rootzén (Chalmers University of Technology, Gothenberg Sweden), Resnick’s Cornell colleague Gennady Samordnitsky, Resnick’s former Ph.D. student Bikramjit Das (Singapore University), and Resnick’s frequent collaborator Richard Davis (Columbia).
Mikosch noted that “Sid is an integral part of the Extreme Value Analysis community,” a group of mathematicians, statisticians and probabilists who work on modelling and understanding the characteristics of phenomena in areas as diverse as finance, telecommunications and weather that exhibit a greater likelihood of extreme events than can be predicted by the common “normal” or “Gaussian” bell-curve probability distribution. A subfamily of probability distributions of such phenomena are sometimes called “heavy-tailed” since there is more area under the tail ends of the plotted curve than is the case for the normal distribution. “I am a great admirer and consider him as one of the godfathers of modern extreme value theory,” said Mikosch, noting that Resnick has also worked with great success “in many other fields of applied probability theory.” In his Sidposium talk, Mikosch spoke on certain mathematical characteristics (eigenvalues and eigenvectors) of matrices associated with heavy-tailed time series distributions.
Rootzén began his talk by referring to a 1977 paper by Resnick and Laurens de Haan. Their paper initiated a whole new research area: modelling and analyzing extreme value properties of multiple interrelated random variables, such as stock prices, communications traffic over multiple links in a network, or rainfall on different days at various points in a river system. Rootzen pointed out that in today’s world real problems may have 50 or more variables, creating a challenge for computation and for the construction of probability models based on these seminal ideas. He proposed new approaches to meeting these challenges so that the techniques can be employed in the analysis of droughts, heat waves, and other extreme events in weather and elsewhere, by modelling multidimensional events that exceed certain thresholds.
At the Sidposium, Davis, who had also participated in the Ithaca Roast, discussed the problem of detecting changepoints and outliers in a series of measurements taken over time. A changepoint might reflect a change in the level of or smoothness of a time series, while outliers indicate suspicious measurement values that appear vastly different than the others. Davis used a concept from information coding theory, called Minimum Description Length (MDL), that is based on determining the minimum amount of memory (bits) required to encode the data. He illustrated his approach with an analysis of data from Google Trends, a tool that records how the frequency with which specific words or terms are used in searches evolves over time. He chose the phrase “Tea Party” during a ten year period from 2004 to 2014, showing a change in the use of the term three weeks after President Obama took office, another change possibly related to news that the IRS was investigating certain political organizations, and outliers occurring when taxes are due. In the spirit of the earlier roast, Davis then applied his methodology to a Google Trends data about searches on “Richard Davis” and attempted to relate changes and outliers to events in his interactions with Resnick.
Samorodnitsky discussed the properties of a particular class of probability distributions, called subexponential distributions, which form “arguably the most general family of models with heavy tails.” While the theory for single variable subexponential distributions is well-established, he noted that it has been difficult to extend the approach to “a useful notion of multivariate subexponentiality.” The single variable version of subexponential distributions has been used in determining the possibility that claims against an insurance company may exceed the accumulated reserves from premiums, leading to ruin. Samorodnitsky demonstrated a model in which his multivariate generalization can be applied to the situation in which an insurance company is in multiple lines of business that may or may not be able to support each other even if a hurricane or earthquake impacts more than one of them. Previous models for this situation have assumed that insurance claims against such a company follow so-called “regularly-varying” probability distributions, but subexponential distributions may not be regularly varying, so this model has broader applicability.
Embrechts continued on the theme of insurance risk, noting that in recent years insurance mathematics has been “a wonderful source of interesting mathematical questions.” He presented results related to a question, brought to him by actuaries at the Zurich Insurance Group, that arose in performing ‘stress testing’ required by insurance regulators. Mathematically, their question in the end boiled down to determining whether a symmetric matrix of numbers in the range from 0 to 1 can interpreted as a matrix of “asymptotic tail dependence coefficients”. If so, these coefficients would describe the tendency of dependence between random variables at the level of their extreme values. Examples where these coefficients are applied are the description of spillover events in financial markets, or in computing the probability that bonds issued to insure against multiple risks (multi-trigger catastrophe bonds) end up losing value for investors because events have occurred that entitle the issuer to avoid payment of interest and repayment of principal. He characterized such tail-dependence matrices, and showed how stochastic models leading to a specific given tail-dependence matrix can be constructed.
Klüppelberg devoted her talk to work by Bruce M. Brown and Resnick on the maximum over a collection of random processes over time of a type, closely related to the “random walk” process, that was originally modelled by Ornstein and Uhlenbeck. In their 1977 paper, Brown and Resnick defined what has become known as the Brown-Resnick Process. Klüppelberg showed how their work extends to encompass not only time, but also space. She has developed specific statistical methods to estimate the extent to which space-time data exhibits dependence at extreme values. She applied the methods to fitting an extreme value model to Radar rainfall measurements taken at 15 minute intervals at 3600 points in a 120x120 km area of Florida, and used various graphical procedures to assess the quality of the model. In a counterweight to the roast, she has described Resnick as “an exceptionally nice person with a remarkable understanding of other people.”
Das, who wrote his dissertation under Resnick and held a post-doctoral position with Embrechts, discussed some ramifications of risk contagion, in which high risk values for one component of an entity or system might influence the risk of another. He analyzed two ways used by financial analysts to quantify this influence, Marginal Mean Excess (MME) and Marginal Expected Shortfall (MES). Both are indicators of risk contagion and useful in various applications including financial insurance and systemic risk to the environment and climate risk. For example, the MES of an institution can considered as the expected loss in equity value if the market itself drops by an extreme amount. Das presented a method to estimate values for MME and MES and showed conditions under which severe risk contagion can occur even if extreme values of the components occur almost independently.
Following a day of deep technical discussion, roasting continued in the bank vault at the Sidposium dinner, where slides evoked such themes such as bicycles, basketball, art, notable Resnick rejoinders, and enduring friendships. At left, Resnick demonstrates two of the gifts he received there: a tee shirt from Embrechts based on statistical charts that are used to detect when a manufacturing or other process is operating within control limits, and from Davis - whom Resnick frequently ribs as being too soft-spoken - the Davis Listening Device, with instructions “for best results, keep mouth shut during use.”