Recent Advances In Monte Carlo Methods For Stochastic Control Problems

Sede FGV - Sala 317

Sobre o Evento

  • Quem: Nicolas Langrené
  • Onde: Praia de Botafogo, 190, sala 317
  • Quando: 23 de Junho de 2016 às 16h

Many problems in practice require to make sequential decisions over time under uncertain, changing conditions. This type of problem can be encountered in finance (how to invest given market uncertainties?), insurance (when to buy an annuity given inflation and longevity uncertainties?), energy (when to operate a hydropower plant given blackout risk?), mining (how much to extract under metal price volatility?), agriculture (best crop rotation under weather uncertainty?), and so on. Mathematically, all these problems can be described as stochastic control problems, also known, depending on the research field, as real option problems, stochastic optimisation problems, or dynamic decision making under uncertainty. In each case, a predefined objective function (or target) has to be maximized over a set of controls (or actions) to adjust over time depending on the time-dependent uncertainties (or risk factors) of the problem at hand. Numerically, three approaches dominate the literature: closed-form, partial differential quations methods (PDEs), and simulation-based methods. Closed-form solutions are usually available for the simplest problems involving a few risk factors with simple dynamics. The PDE approach is always an option, but really shines when few risk factors are involved (up to three in practice). The simulation-based approach, from the simplest trees to the more advanced Monte Carlo simulations, is the most versatile and is suited to high dimensional problems with complex dynamics. This talk will focus on the Monte Carlo approach for solving stochastic control problems. I will detailthe so called Regression Monte Carlo algorithm, first introduced for pricing American options in the mathematical finance literature, and which attracted a lot of attention over the past fifteen years. Then, I will describe several extensions and improvements, in particular the control randomization technique, which allows for endogenous risk factors (factors whose dynamics depend on the optimal control), the memory reduction technique, and the fast adaptive local regression technique.

*Texto informado pelo autor. 



A presença é gratuita e não exige confirmação. A FGV não permite a entrada de pessoas vestindo bermuda e/ou chinelos.


Nicolas Langrené 

PhD in stochastic control at University of Paris VII

Nicolas Langrené completed his engineering studies in computer science & applied mathematics at Grenoble. He did his PhD in stochastic control at University of Paris VII under the direction of Prof Pham. The title was "Probabilistic numerical methods for high-dimensional stochastic control and valuation problems in electricity markets". Then in 2014 he joined the Real Option & Financial Risk team at the CSIRO in Melbourne Australia as a research scientist. He works on designing novel probabilistic algorithms for decision-making under uncertainty.


Sede FGV - Sala 317

Praia de Botafogo, 190


Rio de Janeiro


Praia de Botafogo, 190


Rio de Janeiro

High contrast

Esse site usa cookies

Nosso website coleta informações do seu dispositivo e da sua navegação e utiliza tecnologias como cookies para armazená-las e permitir funcionalidades como: melhorar o funcionamento técnico das páginas, mensurar a audiência do website e oferecer produtos e serviços relevantes por meio de anúncios personalizados. Para mais informações, acesse o nosso Aviso de Cookies e o nosso Aviso de Privacidade.