Conservation equations are at the heart of many interesting and important problems. Examples come from physics, chemistry, biology, traffic and many more. Analytically, hyperbolic equations have a beautiful structure due to the existence of characteristics. These provide the possibility of transforming a conservation PDE into a system of ODE and thus greatly reducing the computational effort required to solve such problems. However, even in one dimension, one encounters problems after a short time.

The most obvious difficulty that needs to be dealt with has to do with the creation of shocks, or in other words, the crossing of characteristics. With a particle based method one would like to avoid a situation when one particle overtakes a neighboring one. However, since shocks are inherent to many hyperbolic equations and relevant to the problems that one would like to solve, it would be good not to ``smooth away'' the shock but rather find a good representation of it and a good solution for the offending particles.

In this talk I will present a new particle based method for solving (one dimensional, scalar) conservation law equations. The guiding principle of the method is the conservative property of the underlying equation. The basic method is conservative, entropy decreasing, variation diminishing and exact away from shocks. A recent extension allows solving equations with a source term, and also provides ``exact'' solutions to the PDE. The method compares favorably to other benchmark solvers, for example CLAWPACK, and requires less computation power to reach the same resolution. A few examples will be shown to illustrate the method, with its various extensions. Due to the current limitation to 1D scalar, the main application we are looking at is traffic flow on a large network. Though we still hope to manage to extend the method to either systems or higher dimensions (each of these extensions has its own set of difficulties), I would be happy to discuss further possible applications or suggestions for extensions.

When a fluid layer is heated from below with temperature larger than a certain critical value, the convective motion appears in the fluid. The convection caused by the thermocapillary effect is called Benard-Marangoni heat convection. The thermocapillary effect is the dependence of the surface tension on the temperature. Near a hot spot on a free surface of fluid a thermocapillary tangential stress generates a fluid motion. In this talk the mathematical model system for this convection is explained. The Oberbeck-Boussinesq approximation is used for the system and the upper boundary is a free surface with surface tension which depends on the temperature. After formulating the linearized problem around the conductive state, we derive the resolvent estimates which guarantee the sectorial property. Stationary and Hopf bifurcations (periodic solutions) are proved to exist depending on the parameters (Raylegh and Marangoni numbers).

We present an overview of experiments, numerical simulations, and mathematical analysis of the breakup of a low viscosity drop in a viscous fluid, and consider the role of surface contaminants, or surfactants, on the dynamics near breakup. As part of our study, we address a significant difficulty in the numerical computation of fluid interfaces with soluble surfactant that occurs in the important limit of very large values of bulk Peclet number Pe. At the high values of Pe in typical fluid-surfactant systems, there is a narrow transition layer near the drop surface or interface in which the surfactant concentration varies rapidly, and its gradient at the interface must be determined accurately to find the drop’s dynamics. Accurately resolving the layer is a challenge for traditional numerical methods. We present recent work that uses the narrowness of the layer to develop fast and accurate `hybrid’ numerical methods that incorporate a separate analytical reduction of the dynamics within the transition layer into a full numerical solution of the interfacial free boundary problem.

In this talk, I will present the optimal transport theory on discrete spaces. Various recent developments related to free energy, Fokker-Planck equations, as well as Wasserstein distance on graphs will be presented, some of them are rather surprising. Applications in robotics as well as Schrodinger equation on graphs will be demonstrated.

$\ell_1$ regularization for sparsity has played important role in recent developments in many fields including signal processing, statistics, optimization. The concept of sparsity is usually for the coefficients (i.e., only a small set of coefficients are nonzero) in a well-chosen set of modes (e.g. a basis or dictionary) for representation of the corresponding vectors or functions. In this talk, I will discuss our recent work on a new use of sparsity-promoting techniques to produce “compressed modes/compress plane waves" - modes that are sparse and localized in space - for efficient solutions of constrained variational problems in mathematics and physics. In particularly, I will focus on L1 regularized variational Schrodinger equations for creating spatially localized modes and orthonormal basis, which can efficiently represent localized functions and has promising potential to a variety of applications in many fields such as signal processing, solid state physics, materials science, etc. (This is a joint work with Vidvuds Ozolins, Russel Caflisch and Stanley Osher)

A new numerical method being developed with Charles Peskin is described which simulates interacting fluid, membrane, and particle systems in which thermal fluctuations play an important role. This method builds on the "Immersed Boundary Method" of Peskin and McQueen, which simplifies the coupling between the fluid and the immersed particles and membranes in such a way as to avoid complex boundary problems. Thermal fluctuations are introduced in the fluid through the theory of statistical hydrodynamics. We discuss some approximate analytical calculations which indicate that the immersed particles should exhibit some physically correct properties of Brownian motion. Our intended use of this numerical method is to simulate microphysiological processes; one advantage this method would have over Langevin particle dynamics approaches is its explicit tracking of the role of the fluid dynamics.

The idea of using nonlinear approximations as a tool for solving equations is as natural as that of using bases and, in fact, was proposed in 1960 in the context of quantum chemistry. The usual approach to solving partial differential and integral equations is to select a basis (possibly a multiresolution basis) or a grid, project equations onto such basis and solve the resulting discrete equations. The nonlinear alternative is to look for the solution within a large lass of functions (larger than any basis) by constructing optimal or near optimal approximations at every step of an algorithm for solving the equations. While this approach can theoretically be very efficient, the difficulties of constructing optimal approximations prevented any significant use of it in practice. However, during the last 15 years, nonlinear approximations have been successfully used to approximate operator kernels via exponentials or Gaussians to any user-specified accuracy, thus enabling a number of multidimensional multiresolution algorithms. In a new development several years ago, we constructed a fast and accurate reduction algorithm for optimal approximation of functions via exponentials or Gaussians (or, in a dual form, by rational functions) than can be used for solving partial differential and integral equations equations. We present two examples of the resulting solvers: one for the viscous Burgers' equation and another for solving the Hartree-Fock equations of quantum chemistry. Burgers' equation is often used as a testbed for numerical methods: if the viscosity \vu; is small, its solutions develop sharp (moving) transition regions of width O (\vu) presenting significant challenges for numerical methods. Using nonlinear approximations for solving the Hartree-Fock equations is the first step to a wider use of the approach in quantum chemistry. We maintain a functional form for the spatial orbitals as a linear combinations of products of decaying exponentials and spherical harmonics entered at the nuclear cusps. While such representations are similar to the classial Slater-type orbitals, in the course of computation we optimize both the exponents and the coefficients in order to achieve an efficient representation of solutions and to obtain guaranteed error bounds.

Coupling between free fluid flow and flow through porous media is important in many industrial applications, such as filtration, underground water flow in hydrology, oil recovery in petroleum engineering, fluid flow through body tissues in biology, to name a few.

Stokes flows appear in many applications where the fluid viscosity is high and/or the velocity and length scales are small. The flow through a porous medium can be described by Darcy's law. A region that contains both requires a careful coupling of these different systems at the interface through appropriate boundary conditions.

Our objective is to develop a method based on the boundary integral formulation for computing the fluid/porous medium problem with higher accuracy using fundamental solutions of Stokes and Darcy's equations. We regularize the kernels to remove the singularity for stability of numerical calculations and eliminate the largest error for higher accuracy.

Motion by mean curvature for a network of surfaces arises in many applications. An important example is the evolution of microstructure in a polycrystalline material under heat treatment. Most metals and ceramics are of this type: They consist of many small single-crystal pieces of differing orientation, called grains, that are stuck together. A famous model proposed by Mullins in the 60s describes the dynamics of the network of surfaces that separate neighboring grains from one another in such a material as gradient descent for a weighted sum of the (possibly anisotropic) areas of the surfaces. The resulting dynamics is motion by weighted mean curvature for the surfaces in the network, together with certain conditions that need to be satisfied at junctions along which three or more surfaces may intersect. Typically, many topological changes occur during the evolution, as grains shrink and disappear, pinch off, or junctions collide. A very elegant algorithm -- known as threshold dynamics -- for the motion by mean curvature of a surface was given by Merriman, Bence, and Osher: It generates the whole evolution simply by alternating two very simple operations: convolution with a Gaussian kernel, and thresholding. It also works for networks, provided that all surfaces in the network have isotropic surface energies with equal weights. Its correct extension to the more general setting of unequal weights and possibly anisotropic (normal dependent) surface energies remained elusive, despite keen interest in this setting from materials scientists. In joint work with Felix Otto, we give a variational formulation of the original threshold dynamics algorithm by identifying a Lyapunov functional for it. In turn, the variational formulation shows how to extend the algorithm correctly to the more general settings that are of interest for materials scientists (joint work with Felix Otto and Matt Elsey). Examples of how to use the new algorithms to investigate unsettled questions about grain size distribution and its evolution will also be given.

In 1953, the physicist P.E. Rouse proposed to model polymers in dilute solution by taking the polymer to be a series of beads connected by Gaussian springs. Neglecting inertia, the dynamics are set by a balance between the thermal fluctuations in the fluid and the elastic restoring force of the springs. One year later, B. Zimm noted that a polymer will interact with itself through the fluid in a qualitatively meaningful way. In this talk, we consider a more recent Langevin equation approach to dealing with hydrodynamic self-interaction. This involves coupling the continuum scaling limit of the Rouse model with stochastically forced time-dependent Stokes equations. The resulting pair of parabolic SPDE, with non-linear coupled forcing, presents a number of mathematical challenges. On the way to providing an existence and uniqueness result, we shall take time to develop relevant stochastic tools, and consider the modeling implications of certain technical results.

In this talk, I will discuss several problems on nonlinear wave and dispersive equations with random initial data, including the energy critical nonlinear wave and Schroedinger equations, and derivative nonlinear wave equations. I will present several almost sure well-posedness and scattering results for these equations and contrast the ways in which random data techniques can be exploited in these different contexts.

Given a number of labeled and unlabeled images, it is possible to determine the class membership of each unlabeled image by creating a sequence of such image transformations that connect it, through other unlabeled images, to a labeled image. In order to measure the total transformation, a robust and reliable metric of the path length is proposed, which combines a local dissimilarity between consecutive images along the path with a global connectivity-based metric. For the local dissimilarity we use a symmetrized version of the zero-order image deformation model (IDM) proposed by Keysers et al. For the global distance we use a connectivity-based metric proposed by Chapelle and Zien in [2]. Experimental results on the MNIST benchmark indicate that the proposed classifier out-performs current state-of-the-art techniques, especially when very few labeled patterns are available.

The Hilbert space dimension of quantum-many body systems grows exponentially with the system size. Fortunately, nature does usually not explore this monstrous number of degrees of freedom and we have a chance to describe quantum systems with much smaller sets of effective degrees of freedom. A very precise description for systems with one spatial dimension is based on so-called matrix product states (MPS). With such a reduced parametrization, the computation cost, needed to achieve a certain accuracy, is determined by entanglement properties (quantum non-locality) in the system.
I will give a short introduction to the notion of entanglement entropies and their scaling behavior in typical many-body systems. I will then employ entanglement entropies to bound the required computation costs in MPS simulations. This will lead us to the amazing conclusion that 1D quantum many-body systems can usually be simulated efficiently on classical computers, both for zero and finite temperatures, and for both gapless and critical systems.
In these considerations, we will encounter a number of mathematical concepts such as the theorem of typical sequences (central limit theorem), concentration of measure (Levy's lemma), singular value decomposition, path integrals, and conformal invariance.

Nonlocal continuum models are in general integro-differential equations in place of the conventional partial differential equations. While nonlocal models show their effectiveness in modeling a number of anomalous and singular processes in physics and material sciences, they also come with increased difficulty in numerical analysis with nonlocality involved. In the first part of this talk, I will discuss nonlocal-to-local coupling techniques so as to improve the computational efficiency of using nonlocal models. This also motivates the development of new mathematical results -- for instance, a new trace theorem that extends the classical results. In the second part of this talk, I will describe our recent effort in computing a nonlocal interface problem arising from segregation of two species with high competition. One species moves according to the classical diffusion and the other adopts a nonlocal strategy. A novel iterative scheme will be presented that constructs a sequence of supersolutions shown to be convergent to the viscosity solution of the interface problem.

Crystals play a critical role in the design of novel devices. The surface of a crystal can evolve with time and give rise to a variety of interesting structures used in applications. In this evolution, several length and time scales, from the atomistic to the continuum, are implicated. The description of their linkages poses challenging questions. How can surface evolution at large scales, possibly described by PDEs, emerge from the motion of mesoscale crystal defects? And how does the description of such defects arise from atomistic motion? In this talk, I will address recent progress and open challenges in answering these questions. In particular, I will discuss how facets, macroscopically flat surface regions, lead to PDE free boundary problems with nontrivial microstructures.

We apply piecewise constructions and gluing technics to construct asymptotically flat (hyperbolic) manifolds such that associated Laplacians have dense embedded eigenvalues or singular continuous spectra. The method also allows us to provide various examples of operators with embedded singular spectra, including perturbed periodic operators, periodic Jacobi operators, and Stark operators. We establish the asymptotic behavior (WKB for example) of eigensolutions under small perturbations, which implies certain rules for the absence of singular spectra. As a result, several sharp spectral transitions (even criteria) for a single (finitely many or countably many) embedded eigenvalues, singular continuous spectra and essential supports of spectral measures are obtained. The talk is based on several papers, some joint with Jitomirskaya and Ong.

I will discuss transport of passive scalars by incompressible flows (such as a die in a fluid) and measures of optimal mixing and stirring under physical constraint on the flow. In particular, I will present recent results concerning examples of flows that achieve the optimal theoretical rate in the case of flows with a prescribed bound on certain Sobolev norms of the associated velocity, such as under an energy or an enstrophy budget. These examples are related to examples of (instantaneous) loss of Sobolev regularity for solutions to linear transport equations with non-Lipschitz velocity.

We focus on first order interacting particle systems, which can be viewed as overdamped Langevin equations. In the first part, we will look at the so-called random batch methods (RBM) for simulating the interacting particle systems. The algorithms are motivated by the mini-batch idea in machine learning. For some special cases, we show the convergence of RBMs for the first marginal under Wasserstein distance. In the second part, we look at the Coulomb interaction in 3D space. We show that as the number of particles go to infinity, almost surely, the empirical measure converges in law to weak solutions of the limiting nonlinear Fokker-Planck equation. This talk is based on joint works with Shi Jin (Shanghai Jiao Tong), Jian-Guo Liu (Duke University) and Pu Yu (Peking University).

It is a common expectation in chemistry that a chemical transformation which takes place in the presence of a catalyst must also take place in its absence, though perhaps at a much slower rate. A reaction network will be called ``saturated'' if it satisfies such an expectation. I propose a mathematical definition for saturated networks and show that the associated dynamical systems have no boundary equilibria in positive stoichiometric classes, and are therefore permanent. This result is independent of the specific rates, and generalizes previous results for complete networks by Gnacadja, atomic event-systems by Adleman et al. and constructive networks by Shinar et al. I require no assumption of complex balance or deficiency restrictions. The question of permanence for weakly-reversible reaction networks remains a long-standing open problem.

The usual pseudospectrum acquires an additional feature when restricted to matrices with a certain symmetry. The new feature is a simple form of K-theory which can be used to compute the index of some one-dimensional topological insulators. The usual pseudospectrum applies to a single matrix, or equivalently to two Hermitian matrices. Generalized to apply to more Hermitian matrices, the nature of the pseudospectrum changes radically, often having interesting geometry. Examples come from D-branes and higher-dimensional topological insulators. The algorithm to compute the pseudospectrum also produces common approximate eigenvectors for a collection of almost commuting Hermitian matrices. Applied to a basic model of a finite volume topological insulator it produces vectors that are approximately stationary and somewhat localized in position.

The Boltzmann equation governs the dynamics of a dilute gas. In this talk, I will address the L^p-stability problem of the Boltzmann equation near vacuum and a global Maxwellian. In a close-to-vacuum regime, I will explain the nonlinear functional approach motivated by Glimm's theory in hyperbolic conservation laws. This functional approach yields the uniform L^1-stability estimate. In contrast, in a close-to-global maxwellian regime, I will present the L^2-stability theory which establishes the uniform L^2-stability of several classical solutions.

Matroids are combinatorial objects that model various types of independence. They appear several fields mathematics, including graph theory, combinatorial optimization, and algebraic geometry. In this talk, I will introduce the theory of matroids along with the closely related class of polynomials called strongly log-concave polynomials. Strong log-concavity is a functional property of a real multivariate polynomial that translates to useful conditions on its coefficients. Discrete probability distributions defined by these coefficients inherit several of these nice properties. I will discuss the beautiful real and combinatorial geometry underlying these polynomials and describe applications to random walks on the faces of simplicial complexes. Consequences include proofs of Mason's conjecture that the sequence of numbers of independent sets of a matroid is ultra log-concave and the Mihail-Vazirani conjecture that the basis exchange graph of a matroid has expansion at least one. This is based on joint work with Nima Anari, Kuikui Liu, and Shayan Oveis Gharan.

With Jon Weare (Chicago), we study the continuum limit of a family of kinetic Monte Carlo models of crystal surface relaxation that includes both the solid-on-solid and discrete Gaussian models. With computational experiments and theoretical arguments we are able to derive several partial differential equation limits identified (or nearly identified) in previous studies and to clarify the correct choice of surface tension appearing in the PDE and the correct scaling regime giving rise to each PDE. We also provide preliminary computational and analytic investigations of a number of interesting qualitative features of the large scale behavior of the models. The PDE models involved are fully non-linear Fourth order diffusion type equations with many interesting geometric features. We will given time discuss recent progress analyzing properties of solutions to such PDE.

We survey recent advances on the complexity and methods for solving Markov decision problems (MDP) and Reinforcement Learning (RL) with finitely many states and actions - a basic mathematical model for reinforcement learning.

For model reduction of large scale MDP in reinforcement learning, we propose a bilinear primal-dual pi learning method that utilizes given state and action features. The method is motivated from a saddle point formulation of the Bellman equation. The sample complexity of bilinear pi learning depends only on the number of parameters and is variant with respect to the dimension of the problem.

In the second part we study the statistical state compression of general Markov processes. We propose a spectral state compression method for learning the state features from data. The state compression method is able to “ sketch” a black-box Markov process from its empirical data and output state features, for which we provide both minimax statistical guarantees and scalable computational tools.

We will discuss some recent results on singularity formation for finite-energy strong solutions to the 3D Euler system based on the analysis of scale-invariant solutions. The work consists of three parts: local well-posedness in critical spaces, the proof of blow-up for scale-invariant solutions, and then a cut-off argument to ensure finite energy. This is joint work with I. Jeong.

Multi-segment reconstruction (MSR) problem consists of recovering a signal from noisy segments with unknown positions of the observation windows. One example arises in DNA sequence assembly, which is typically solved by matching short reads to form longer sequences. Instead of trying to locate the segment within the sequence through pair-wise matching, we propose a new approach that uses shift-invariant features to estimate both the underlying signal and the distribution of the positions of the segments. Using the invariant features, we formulate the problem as a constrained nonlinear least-squares. The non-convexity of the problem leads to its sensitivity to the initialization. However, with clean data, we show empirically that for longer segment lengths, random initialization achieves exact recovery. Furthermore, we compare the performance of our approach to the results of expectation maximization and demonstrate that the new approach is robust to noise and computationally more efficient.

Recently there has been a large class of research that utilizes nonlinear mappings into low dimensional spaces in order to organize potentially high dimensional data. Examples include, but are not limited to, locally linear embedding (LLE), ISOMAP, Hessian LLE, Laplacian eigenmaps, and diffusion maps. In this talk we will focus on the latter, and in particular consider how to generalize diffusion maps to the setting in which we are given a data set that evolves over time or changes depending on some set of parameters. Along with describing the current theory, various synthetic and real world examples will be presented to illustrate these ideas in practice.

Exploring data structures (e.g, periodicity, sparsity, low-rankness) is a universal method in designing fast algorithms in scientific computing. In the first part of this talk, I will show how this idea is applied to the analysis of oscillatory data in applied harmonic analysis. These fast algorithms have been applied to data analysis ranging from materials science, medicine, and art. In the second part, I will discuss how this idea works in some basic numerical linear algebra routines like matrix multiplications and decompositions, with an emphasis in electronic structure calculation.

Fluid-structure interactions exist in many aspects of our daily lives. Some biomedical engineering examples are blood flowing through a blood vessel and blood pumping in the heart. Fluid interacting with moving or deformable structures poses more numerical challenges for its complexity in dealing with transient and simultaneous interactions between the fluid and solid domains. To obtain stable, effective, and accurate solutions is not trivial. Traditional methods that are available in commercial software often generate numerical instabilities.

In this talk, a novel numerical solution technique, Immersed Finite Element Method (IFEM), is introduced for solving complex fluid-structure interaction problems in various engineering fields. The fluid and solid domains are fully coupled, thus yield accurate and stable solutions. The variables in the two domains are interpolated via a delta function that enables the use of non-uniform grids in the fluid domain, which allows the use of arbitrary geometry shapes and boundary conditions. This method extends the capabilities and flexibilities in solving various biomedical, traditional mechanical, and aerospace engineering problems with detailed and realistic mechanics analysis. Verification problems will be shown to validate the accuracy and effectiveness of this numerical approach. Several biomechanical problems will be presented: 1) blood flow in the left atrium and left atrial appendage which is the main source of blood in patients with atrial fibrillation. The function of the appendage is determined through fluid-structure interaction analysis, 2) examine blood cell and cell interactions under different flow shear rates. The formation of the cell aggregates can be predicted when given a physiologic shear rate.