We propose a novel approach of numerically approximate McKean-Vlasov SDEs that avoids the usual interacting particle approximation and Propagation of Chaos results altogether.

The aim of this talk is to investigate the following non local p-Laplacian problem with data a bounded Radon measure.

Applied Complexity aims to understand the physical origin of these behaviors and transform them into sustainable technologies that tackle global problems of global interest. These range from energy harvesting to clean water production, the design of smart materials, biomedical applications, information security, artificial intelligence, and global warming. In this talk, I will summarize my group's recent research, discussing present results and future challenges of Applied complexity both as a science and engineering.

Models for flows on networks arise in the study of traffic and pedestrian crowds. These models encode congestion effects, the behavior and preferences of agents, such as aversion to crowds and their attempts to minimize travel time. We will present the Wardrop equilibrium model on networks with flow-dependent costs and its connection with stationary mean-field game.

In this work, we proved statistical inference for the common Z-estimator based on adaptively sampled data. Adaptive sampling methods, such as reinforcement learning (RL) and bandit algorithms, are increasingly used for the real-time personalization of interventions in digital applications like mobile health and education. As a result, there is a need to be able to use the resulting adaptively collected user data to address a variety of inferential questions, including questions about time-varying causal effects.

Reinforcement Learning provides an attractive suite of online learning methods for personalizing interventions in Digital Behavioral Health. However, after a reinforcement learning algorithm has been run in a clinical study, how do we assess whether personalization occurred? We might find users for whom it appears that the algorithm has indeed learned in which contexts the user is more responsive to a particular intervention. But could this have happened completely by chance? We discuss some first approaches to addressing these questions.

In the Seminar, we exploit what is the heart of the technique to show gradient estimates allowing the function 𝑓 to belong to very general function spaces. Our technique is very flexible and allows us to show the existence, uniqueness, and well-posedness of the Dirichlet problem in several classes.

In this talk, we show that, besides their optimal O(N) algorithmic complexity, hierarchical matrix operations also benefit from parallel scalability on distributed machines with extremely large core counts. In particular, we describe high-performance, distributed-memory, GPU-accelerated algorithms for matrix-vector multiplication and other operations on hierarchical matrices in the H^2 format.

This talk discusses the non-conforming approximation of Biot's consolidation model. In the first part of the talk, we discuss posteriori error estimators for locking-free mixed finite element approximation of Biot’s consolidation model. In the second part of the talk, we discuss a novel locking-free stochastic Galerkin mixed finite element method for the Biot consolidation model with uncertain Young’s modulus and hydraulic conductivity field.

We present a new constrained K-way data clustering algorithm based on Normalized Cut by Shi and Malik. A novelty in our algorithm lies in selecting constraints automatically from the data by using a multiscale coarsening algorithm.

Eigenvalue problems arising from partial differential equations are used to model several applications in science and engineering, ranging from vibrations of structures, industrial microwaves, photonic crystals, and waveguides, to particle accelerators.

Freeform structures play a prominent role in contemporary architecture. In order to stay within reasonable cost limits, computational shape design has to incorporate aspects of structural analysis and fabrication constraints. The talk discusses solutions to important problems in this area. They concern the design of polyhedral surfaces with nearly rectangular faces, polyhedral surfaces in static equilibrium, the smoothest visual appearance of polyhedral surfaces and the closely related problem of finding material-minimizing forms and structures. From a methodology perspective, there is an interplay of geometry, mechanics and optimization. Classical subjects such as isotropic geometry, a simple Cayley-Klein geometry, play a role as well as most recent developments in discrete differential geometry. We also show how practical requirements have led to new results and open problems in geometry.

Dynamic programming is an efficient technique to solve optimization problems. It is based on decomposing the initial problem into simpler ones and solving these sub-problems beginning from the simplest ones. A conventional dynamic programming algorithm returns an optimal object from a given set of objects.

Random fields are popular models in statistics and machine learning for spatially dependent data on Euclidian domains. However, in many applications, data is observed on non-Euclidian domains such as street networks. In this case, it is much more difficult to construct valid random field models. In this talk, we discuss some recent approaches to modeling data in this setting, and in particular define a new class of Gaussian processes on compact metric graphs.

An efficient method is proposed for the numerical solution of the Stokes equations in a domain with a moving bubble and two techniques for the treatment of the boundary conditions are adopted and then compared. The treatment of diffusion of surfactants (anions and cations) in presence of an oscillating bubble is an interesting interdisciplinary problem, with applications to chemistry and biology.

In this talk, I will first give an elementary introduction to basic deep learning models and training algorithms from a scientific computing viewpoint. Using image classification as an example, I will try to give mathematical explanations of why and how some popular deep learning models such as convolutional neural network (CNN) work. Most of the talk will be assessable to an audience who have basic knowledge of calculus and matrix. Toward the end of the talk, I will touch upon some advanced topics to demonstrate the potential of new mathematical insights for helping understand and improve the efficiency of deep learning technologies.

In this paper, we propose a detailed theoretical study of one of these algorithms known as the split Gibbs sampler. Under regularity conditions, we establish explicit convergence rates for this scheme using Ricci curvature and coupling ideas. We support our theory with numerical illustrations.

No abstract is available.

In this talk, I will explain the problem, its solution, and some subsequent work generalizing, extending and improving the ProxSkip method in various ways. We study distributed optimization methods based on the local training (LT) paradigm - achieving improved communication efficiency by performing richer local gradient-based training on the clients before parameter averaging - which is of key importance in federated learning. Looking back at the progress of the field in the last decade, we identify 5 generations of LT methods: 1) heuristic, 2) homogeneous, 3) sublinear, 4) linear, and 5) accelerated. The 5th generation, initiated by the ProxSkip method of Mishchenko et al (2022) and its analysis, is characterized by the first theoretical confirmation that LT is a communication acceleration mechanism.

No abstract is available.

Abstract

Two or multiple phases in fluid mixture commonly occur in petroleum industry, where oil,

Tile low-rank and hierarchical low-rank matrices can exploit the data sparsity that is discoverable all across computational science. We illustrate in large-scale applications and hybridize with similarly motivated mixed precision representations while featuring ECRC research in progress with many collaborators.

In this talk, we shall explain how transportation networks emerge as a self-regulating process, with a particular focus on applications in biology (leaf venation in plants, neuronal networks in animals). We start by introducing a purely diffusive model with tensor-valued diffusivity, derived as a gradient flow of a broad class of entropy dissipations. The introduction of a prescribed electric potential leads to the Fokker-Planck equation. We show that with quadratic entropy density modeling Joule heating, the model is convex with respect to the diffusivity tensor.

In this section, prof. Gabriel will be introducing the graduate seminar and its requirements.

This thesis consists of three main parts. In the first part, we discuss first-order stationary mean-field games (MFGs) on networks. In the second part, we discuss the Wardrop equilibrium model on networks with flow-dependent costs and its connection with stationary MFGs. First, we build the Wardrop model on networks. Second, we show how to convert the MFG model into a Wardrop model. Next, we recover the MFG solution from the Wardrop solution. Finally, we study the calibration of MFGs with Wardrop travel cost problems. In the third part, we explain the algorithm for solving the algebraic system associated with the MFG numerically, then, we present some examples and numerical results.