The 2009 annual meeting of the Society for Industrial and Applied Mathematics (SIAM) drew approximately 950 attendees to the Colorado Convention Center, July 610. (The event had been scheduled for the Denver Sheraton, but construction at the hotel forced a relocation. Even though the people at SIAM weren't notified until about two weeks before the meeting, the relocation came off smoothly.)
Below are links to summaries of some of the events that took place at the meeting.
* Kill All the Quants? * The H1N1 Flu * Prizes and Awards Luncheon * Cleve Moler's Past President's Address * Optimization: The Difference Between Theory and Practice * Energy and Earth Systems Modeling * Modeling Cancer * Convection in the Earth's Mantle * John von Neumann Lecture * Bias in Standardized Tests * Exhibits
"Kill All the Quants?": Models vs. Mania in the Current Financial Crisis, the I.E. Block Community Lecture, Andrew Lo, Massachusetts Insitute of Technology
Lo said that it was a rare privilege for him, an economist, to speak to an audience of mathematicians. From the audience's point of view it was also a privilege to hear this lecture in which he explained the basics of some of the financial instruments in the news and gave recommendations for how to reduce the damage from future financial catastrophes. The answer to the question in the lecture title is No: Lo said that blaming quantitative analysts for the financial crisis "is like blaming arithmetic for accounting fraud." He showed a graph (at left) of real home prices over the years and noted the huge spike before the meltdown. Although home prices are now down, they haven't yet returned to the level before the upsurge that began in the 1990s.

The basic example that he gave of a financial instrument involved two IOU's each for $1000 and each with a 10% risk of default. This is a high risk, so most banks would avoid buying them. But now create two securities: one blue and the other orange with the stipulation that the blue security will pay off before the orange one. Then assuming independence of the two defaults, there is a 99% chance that the blue security will pay off, since both 10% defaults would have to occur. So now, in Lo's words, "Two ugly parents [the original IOU's] have created one very attractive offspring [blue] and one ugly one [orange]." Note that risktakers will still buy the orange security. In any case, as long as the defaults are independent, the blue security is a good risk, but given the interconnectedness of markets, the independence assumption is not justified and so what was supposed to be a safe investment is in fact a risky one. This is a simplified example of what has happened globally.
Lo said that in general technologybased industries, such as finance, "accidents" happen all the time. Charles Perrow (Normal Accidents: Living with HighRisk Technologies) attributes this to the complexity or nonlinearity of systems as well as to the tight coupling of procedures, in which each step must follow a previous step. To these conditions, Lo added the absence of negative feedback over an extended period of time. He recommended breaking up banks that are "too big to fail," requiring certification for those who manage complex financial instruments, and teaching economics and finance in high schools.
Mathematics of Influenza, Mac Hyman, Los Alamos National Laboratory
Hyman gave an overview of H1N1 and other flu viruses, and how the spread of flu is modeled. In the standard susceptibleinfectedrecovered (SIR) model, almost everyone is susceptible to H1N1 because we currently have no immunity against it, which means many people could be infected. So far, although H1N1 is highly transmissible, it is mild. The H1N1 virus is different from seasonal flus, and is made up of one piece from birds, two from pigs, and one from humans. This means that the usual threecomponent seasonal flu vaccine isn't effective. As for the progression of the disease, people don't notice symptoms in the first three days of being infected but may feel better after the fifth dayand so be tempted to return to work or schooleven though they are still contagious. To minimize your chance of getting the flu, Hyman advises keeping a social distance of 36 feet, and washing your hands. He said masks only keep those who have the H1N1 virus from transmitting it to othersmasks worn by those who are uninfected won't keep the flu virus out, but are only effective in that they keep people from touching their faces. [Note: In a later interview done with Mac about the flu (12MB pdf), he talked about a study at the University of Michigan that said that handwashing and masks can cut the rate of transmission in half.] 
SIAM President Doug Arnold hosted this luncheon and handed out the following prizes and awards.
In addition, the first class of SIAM Fellows was recognized. Those in attendance are pictured below. The complete list of the 191 SIAM Fellows is online.
Parallelism and Puzzles, Past President's Address, Cleve Moler, The MathWorks, Inc.
Moler was introduced by SIAM President Doug Arnold who observed that giving the Past President's Address is similar to the sentiment felt by Michael Corleone in The Godfather: Part III: "Just when I thought I was out, they pull me back in." Moler began his address by reflecting on his career and saying that over the years his association with SIAM has been the high point of his career. The first part of his talk was a survey of parallel computing, from its beginning with floating point operations to today with multicore processors and multicomputer parallelism. Moler is a bit dubious about parallelism in computing, however. He noted that "The number of people who say Moore's Law is dead doubles every 18 months." In the second part of the address, he talked about a project of his, Experiments with Matlab, which is for students and those interested in recreational mathematics. 
Optimization: The Difference Between Theory and Practice, Juan C. Meza, Lawrence Berkeley National Laboratory
The calculated dipole moment of a 2633 atom CdSe quantum rod, Cd961Se724H948. The calculation took 30 hours using 2560 processors at the National Energy Research Scientific Computing Center. Juan Meza et al., Lawrence Berkeley National Laboratory. For more information, see http://hpcrd.lbl.gov/~meza 
Optimization is an increasingly widespread application as computational capacity increases. Even though most optimization problems are very difficulty to analyze theoretically, solutions are still required. One of the biggest challenges in simulationbased optimization is the computational cost of objective functions. In fact, Meza said that the dominant cost in most problems is evaluation of the objective function. He then talked about the particular issues in three case studies: chemical vapor deposition control (for making silicon chips), supernovae spectra (used to probe dark energy), and computational nanoscience (to create nanorods for solar cells). Meza concluded his invited topical address by stating that theory gives the framework to analyze problems and guide solutions, while practice gives the experience of realworld problems and helps improve algorithms. 
The Mathematics Behind Energy and Earth Systems Modeling, Juan C. Meza, Lawrence Berkeley National Laboratory
Simulation of a leanpremixed hydrogen flame stabilized on a lowswirl burner. The picture depicts the concentration of OH and vorticity. John Bell et al., Lawrence Berkeley National Laboratory. For more information see https://seesar.lbl.gov/CCSE/ 
Meza also gave a shorter talk at the meeting, this one as part of the Workshop Celebrating Diversity. He first talked about the rise in temperatures and global carbon dioxide levels. Tracing the history of study of the greenhouse effect, Meza gave a quote from 1824 by Joseph Fourier who noted that incoming solar radiation penetrates the atmosphere easier than the reflected infrared rays, so the latter can be trapped. He showed what mathematicians are doing to: describe and predict the climate, improve our use of energy (good simulations of combustion are now possible) and change how we use energy in the figure (especially analyzing how air circulates in buildingsin the U.S. buildings account for 39% of energy use). Meza concluded by saying that mathematics is the foundation of climate and energy models. 
Modeling CancerImmunology Dynamics, Lisette de Pillis, Harvey Mudd College
In this invited topical address, de Pillis explained some basics of cancer cells and work she had done modeling the effect of chemotherapy on the progress of the disease. The research began by trying to answer the question of why tumors sometimes grow when treated with chemotherapy and shrink when not treated. Her answer is tied to a patient's immune system. She and her colleagues have developed models based on ordinary differential equations, that incorporate tumor cells, host cells, immune cells, and drug interaction. The models have been successful keeping patients in a region with a favorable basin of attraction (in which the tumor shrinks) and shifting them there, by gathering data on an individual's immune system and using that data to guide the treatment. The prescribed treatment uses a combination of chemotherapy and immunotherapy. Before her research, patients could be in different basins of attraction and the same chemotherapy treatment would push them further into the region, which is bad for those starting off in the "wrong" region. By knowing the state of a patient's immune system, doctors can identify which region a patient is in, and prescribe treatment based on the region.
Towards FullyImplicit Parallel Adaptive Solution of Mantle Convection Problems, Lucas Wilcox, University of Texas at Austin

Mantle convection is the primary control for heat transfer in the Earth, which makes it important to study to understand plate tectonics, volcanoes, earthquakes, and mountain building. Models in this field use an advectiondiffusion equation and a nonlinear Stokes equation. Wilcox and his colleagues are developing the first global mantle convection simulation that resolves plate boundaries under real conditions. The simulations encompass a wide spatial scaleone kilometer to 10^{4} kmand a large time scale10^{4} years to 10^{9} years. One of the hardest parts of refining the models is obtaining a good approximation of viscosity.

Compatible Discretizations of PDE's, the John von Neumann Lecture, Franco Brezzi, Institute of Applied Mathematics and Information Technology (IMATI) and Institute for Advanced Study (IUSS) (Italy)
Brezzi started out by saying that he could not live up to the nice introduction given by SIAM President Doug Arnold, because he was not actually Franco Brezzi, but rather is twin brother. Franco, who is "very bright," was on vacation and so "they sent his twin." In his talk, he contrasted different discretization methods and the way functions are identifiedeither by assigning parameters in a finitedimensional vector space or through knowledge of values at certain points. The first type of discretization reproduces physical laws in a weak sense, while the second reproduces them in a strong sense. He concluded with an outline of the steps in Mimetic Function Difference methods, in which after a few steps involving nodal functions and inner products, a "miracle" occurs in the final step of filling out a matrix.
Are Standardized Tests Biased Against Minorities?, Steve Culpepper, University of ColoradoDenver
Culpepper reviewed the recent U.S. Supreme Court decision in Ricci v. DeStefano (the New Haven firefighters case) in which the Court ruled that the test in question was fair. He pointed out that "fair" is not an easy concept to define, but approached it from a meritocratic point of view and discussed the statistical tools used to analyze that type of fairness. Culpepper gave theoretical college admissions examples involving standardized test scores and regression lines. He said that when a test is fair for the majority population but not for minorities, and admissions officers truncate the regression lines (which happens because of cutoff scores), it is hard to detect test bias against minorities using standard leastsquares regression techniques. In fact, the techniques are likely to lead to the conclusion that there is bias against the majority in such cases.
Several exhibitors were present at the meeting, including the AMS, and got the most traffic during breaks between lectures when free coffee and tea were available.
Next year's meeting will take place in Pittsburgh, July 1216. A description of the 2008 SIAM annual meeting in San Diego with links to descriptions of previous SIAM annual meetings is also available.
Photos and text by AMS Public Awareness Officer Mike Breen.