- in memoriam: José Rodrigues Dias (1951-2023)
- in memoriam: Vladimir Goncharov (1962-2017)
- in memoriam: Graça Carita (1975-2016)
- Notícias
- Eventos
-
Seminários
- The representation theory of the unitriangular group and other related groups
- On Computational Properties of Cauchy Problems
- Sets as Properties
- Artinian algebras and Jordan type
- Iterated Systems, Networks and Applications
- Isogeometric Analysis: mathematical and implementational aspects, with applications
- Sampling elusive populations: methods and applications
- Existence, non-existence and multiplicity results for some third-order periodic problems
- Convergence: what’s logic got to do with it?
- The Mathematics of Fires
- Old-age mortality deceleration and the modal age at death: insights from dynamic laws of adult mortality
- Stochastic differential equations models of animal growth and profit optimization in cattle raising
- Multidisciplinary approach for a real problem: modeling road traffic accidents
- The finite elements method and Freefem software
- DeepParticle: deep-learning invariant measure by minimizing Wasserstein distance on data generated from an interacting particle
- Kinematics: classification methods and combinatorial invariants for complex motion in biology
- Fractional Poisson Analysis in one Dimension
- Mathematics driven by epidemics
- ALMOST-POSITIONED NUMERICAL SEMIGROUPS
- A Delta approximation method on estimation for SDE mixed models
- Classificação de uma família de nós de Lorenz redutíveis
- Categoria de Lusternik-Schinrelmann e Grupos de Lie
- Euclides, Taylor, e a perspectiva esférica enquanto objecto matemático
- Time Series Clustering
- Condições geométricas para a existência e unicidade de projeção
- Sobre o volume de campos vetoriais
- Multiple criteria optimization: methods and applications
- Selective Base Revisions
- Nonautonomous attractors and bifurcation structures on nonautonomous families of flat topped tent maps
- Strongly nonlinear third order impulsive boundary value problems
- Solvability of second order coupled systems on the half-line
- Stochastic differential equations: brief introduction and profit optimization in fisheries
- Positioned Numerical Semigroups
- Tipo de Jordan de álgebras artinianas
- Complexidade em sistemas dinâmicos de baixa dimensão
- CFD Analysis in cerebral aneurysms
- Variational problems involving nonlocal supremal functionals
- Different Types of Stabilities in Times of Instability
- Non-linear systems of PDEs. Two examples from applications.
- Modelação de eventos extremos – uma introdução: aplicação ao decatlo e ao heptatlo atlético
- Comportamento assimptótico de soluções de problemas com valores na fronteira
- A característica de Euler de hipersuperfícies de espaços forma
- Brief introduction to stochastic differential equations and applications in Biology and Finance
- Sampling strategies in rural and urban settings in Africa - Looking from the sky
- Epidemiologia espácio-temporal no controlo da tuberculose
- Connectivity and Reliability of Mobile Ad-Hoc Networks
- Mathematical modeling, optimal control and complex network of epidemic models: case study of COVID-19 in Portugal
- Feature selection for marine species origin prediction
- Provas de Mestrado
- Provas de Doutoramento
- Apontadores
DeepParticle: deep-learning invariant measure by minimizing Wasserstein distance on data generated from an interacting particle
High dimensional partial differential equations (PDE) are challenging to compute by traditional mesh based methods especially when their solutions have large gradients or concentrations at unknown locations. Mesh free methods are more appealing, however they remain slow and expensive when a long time and resolved computation is necessary.
We present DeepParticle, an integrated deep learning (DL), optimal transport (OT), and interacting particle (IP) approach through a case study of Fisher-Kolmogorov-Petrovsky-Piskunov front speeds in incompressible flows.
PDE analysis reduces the problem to a computation of principal eigenvalue of an advection-diffusion operator. Stochastic representation via Feynman-Kac formula makes possible a genetic interacting particle algorithm that evolves particle distribution to a large time invariant measure from which the front speed is extracted. The invariant measure is parameterized by a physical parameter (the Peclet number). We learn this family of invariant measures by training a physically parameterized deep neural network on affordable data from IP computation at moderate Peclet numbers, then predict at a larger Peclet number when IP computation is expensive.
The network is trained by minimizing a discrete Wasserstein distance from OT theory. The DL prediction serves as a warm start to accelerate IP computation especially for a 3-dimensional time dependent Kolmogorov flow with chaotic streamlines.
Our methodology extends to a more general context of deep-learning stochastic particle dynamics.This is joint work with Zhongjian Wang (University of Chicago) and Zhiwen Zhang (University of Hong Kong).