There is strong evidence that quasinormal modes, eigenmodes of non-self adjoint operators, are important in physical systems. There are equally strong arguments that quasinormal modes cannot have any real significance or useful mathematical characterization. Both sides of this question will be presented without bias.
October 23, 2015
The Climate Corporation
"Green Data Revolution in Agriculture -- Data Science @ The Climate Corporation"
Abstract: With more and more data been collected in modern agriculture, data science is actively reforming the agricultural industry. In this talk, I’ll give an overview of data science at the Climate Corporation, which aims to help farmers sustainably increase their productivity with digital tools. I will cover varieties of changes we’re facing as data scientists in this relative new area and give examples to illustrate how we can apply modern statistical and machine learning techniques in this interdisciplinary area involving remote sensing, plant science, meteorology etc.
Bio: Ying Xu is a Senior Quantitative Researcher at the Climate Corporation (TCC). She received her BS in Mathematics from Peking University, and Ph.D. in Statistics from U.C. Berkeley under the supervision of Prof. Peter Bickel. Her current research interests lie primarily in applying statistical and machine learning techniques to precision agriculture.
"Derivation and Simulation of Nonequilibrium Langevin Dynamics"
Abstract: The project is interested in methods to sample particle systems that have an overall steady, homogeneous flow. One application of this dynamics is to impose a strain rate on a complex fluid or immersed molecular system in order to compute the stress-strain constitutive relation using a microscopic stress formulation. We describe a derivation of a nonequilibrium Langevin dynamics for sampling systems embedded in a steady, non-uniform flow. The equations of motion are the limit of a family of mechanical systems containing a single large particle immersed in an infinite bath of small atoms that have a consistent mean velocity gradient, and we find that the large particle evolves according to a stochastic dynamics. We also discuss algorithmic aspects of nonequilibrium molecular dynamics simulations. In such a simulation the simulation box deforms with the flow, and we describe particular boundary conditions that allow for long-time simulation by avoiding extreme deformation of the unit cell.
"Applications of Mining Heterogeneous Information Networks"
Abstract: Most real-world applications that handle big data, including interconnected social media and social networks, scientific, engineering, or medical information systems, online e-commerce systems, and most database systems, can be structured into heterogeneous information networks. Different from homogeneous information networks, where objects and links are treated either as of the same type or as of untyped nodes or links, heterogeneous information networks in our model are semi-structured and typed, following a network schema. Recent studies have demonstrated the power of heterogeneous information networks in many real-world applications. In this talk, I will introduce some of these interesting studies, which include (1) recommendation, (2) information diffusion, and (3) ideology detection in heterogeneous information networks.
Bio: Yizhou Sun is an assistant professor in the College of Computer and Information Science of Northeastern University. She received her Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign in 2012. Her principal research interest is in mining information and social networks, and more generally in data mining, machine learning, and network science, with a focus on modeling novel problems and proposing scalable algorithms for large-scale, real-world applications. Yizhou has over 60 publications in books, journals, and major conferences. Tutorials based on her thesis work on mining heterogeneous information networks have been given in several premier conferences, including EDBT 2009, SIGMOD 2010, KDD 2010, ICDE 2012, VLDB 2012, and ASONAM 2012. She received 2012 ACM SIGKDD Best Student Paper Award, 2013 ACM SIGKDD Doctoral Dissertation Award, 2013 Yahoo ACE (Academic Career Enhancement) Award, and 2015 NSF CAREER Award.
October 6, 2015
Beijing Computational Science Research Center
"Superconvergence of Discontinuous Galerkin methods based on upwind-biased fluxes for 1D linear hyperbolic equations"
Abstract: In this talk, we analyze the superconvergence properties of discontinuous Galerkin methods using upwind-biased numerical fluxes for one-dimensional linear hyperbolic equations. A (2k+1)th order superconvergence rate of the DG approximation at the numerical fluxes and for the cell average is obtained under quasi-uniform meshes and some suitable initial discretization, when piecewise polynomials of degree $k$ are used. Furthermore, surprisingly, we find that the derivative and function value approximations of the DG solution are superconvergent at a class of special points, with an order (k+1) and (k+2), respectively. These superconvergence points can be regarded as the generalized Radau points.
"FICO: Make Every Decision Count™ Overview of FICO / Advanced Analytics "
Abstract: In this presentation, Dr. Chenyang Lian will first give an overview of FICO, a leading analytic software company that helps business grow by making data-driven decisions. Organizations in 90+ countries turn to FICO to improve customer engagement, automate, and optimize decisions for higher ROI. Dr. Lian will also introduce FICO’s advanced analytics such as decision optimization, and how that has been applied in the real world to help clients make better and smarter decisions across the industries. He will also discuss the latest hot topics on alternative data and big data analytics and how they transform the traditional analytics in the industry.
Southwest University of Science and Technology; China
"Some Integral Inequalities and Their Analogues with Applications"
Abstract: In this talk, some new weakly singular versions of nonlinear integral inequalities and discrete analogues are established, which generalize some existing results and can be used in the analysis of nonlinear Volterra type differential equations and difference equations with weakly singular kernels, respectively. Furthermore, Gronwall-Bellman-Gamidov integral inequality is revised and also generalized to some new cases. Some applications to the upper bound and uniqueness of solutions of some nonlinear equations are also given.
"Algebraic Multigrid Method and Its Parallelization"
Abstract: Developing parallel algorithms for solving large-scale sparse linear systems is an important and challenging task in scientific computing and practical applications. In this talk, I will introduce the unsmoothed aggregation algebraic multigrid method for solving large-scale linear systems. I will give theoretical justifications of its optimality for model problems and its parallelization, especially on graphic processing units. Two different parallel approaches will be discussed and numerical results will be presented to demonstrate their efficiency.
Abstract: A fundamental idea in matrix linear algebra is the factorization of a matrix into simpler matrices, such as orthogonal, tridiagonal, and triangular. In this talk we extend this idea to a continuous setting, asking: "What are the continuous analogues of matrix factorizations?" The answer we develop involves functions of two variables, an iterative variant of Gaussian elimination, and sufficient conditions for convergence. This leads to a test for non-negative definite kernels, a continuous definition of a triangular quasimatrix (a matrix whose columns are functions), and a fresh perspective on a classic subject.
BlueMountain Capital Management; New York (Previously: Morgan Stanley; Massachusetts Institute of Technology)
"Careers in Quantitative Finance for Math"
Abstract: Science, and Engineering graduates, Investment banks, hedge funds, and asset managers routinely fill their ranks with recent science, engineering, and math graduates. I will present why such a need exists, what a career in quantitative finance entails, and how you can successfully prepare for one. This talk will draw from my experiences both as a physics grad student and a working professional in the industry.
This seminar is sponsored jointly by the Engineering and Applied Sciences program at UMassD.
"Coarse-graining high-dimensional random systems: the probability density function approach"
Abstract: In this talk, I will present different methods to develop reduced-order equations for the probability density function (PDF) of quantities of interest (phase space functions) in high-dimensional stochastic dynamical systems. In principle, this allows us to avoid computing high-dimensional stochastic flows and reduce a great mass of information to a more tractable and more interesting form (the PDF evolution of a low-dimensional quantity of interest). I will discuss two different classes of methods to achieve this goal: the first one is based on separated series expansions such as proper generalized decomposition (PGD) or high-dimensional model representations (HDMR). When applied to high-dimensional systems, these methods yield a hierarchy of low-dimensional PDF equations that resembles the Bogoliubov-Born-Green-Kirkwood-Yvon (BBGKY) hierarchy of kinetic gas theory. The second approach stems from techniques of irreversible statistical mechanics, in particular the Mori-Zwanzig (MZ) formulation, and it yields directly formally exact equations for the PDF of quantities of interest. I will address the question of approximation of MZ-PDF equations by multi-level coarse graining, perturbation series and operator cumulant resummation. Throughout the presentation I will provide numerical examples and applications of the proposed methods to prototype stochastic problems, such as the Lorenz-96 system, stochastic advection-reaction and stochastic Burgers equations.
"Optimal error estimate of spectral Galerkin and collocation methods for fractional differential equations"
Abstract: We present optimal error estimates for spectral Galerkin methods and spectral collocation methods for linear fractional differential equations with initial value or boundary values on a finite interval. The key feature of the error estimates is that the estimate can accommodate the end-point singularity of the solutions to these equations. We also develop Laguerre spectral Petrov-Galerkin methods and collocation methods for fractional equations on the half line. Numerical results confirm the error estimates.
Abstract: http://www.durkheimproject.org The project is named in honor of Emile Durkheim, a founding sociologist whose 1897 publication of Suicide defined early text analysis for suicide risk, and provided important theoretical explanations relating to societal disconnection. The Durkheim Project is comprised of a multidisciplinary team of artificial intelligence (machine learning) and medical experts (psychiatrists) from Dartmouth Engineering, Dartmouth Medical School, and the U.S. Veterans Administration. Together these professionals have formed a team dedicated to applied research on predictive suicide risk.