14.10.14
Raphael Kruse (TU Berlin)
Numerical solution of stochastic differential equations under one-sided Lipschitz conditions
Abstract: In this talk we present a new method of proof which yields the strong order of convergence for Euler-type numerical approximations of stochastic differential equations. Hereby we assume that the drift and diffusion coefficient functions fulfill a suitable one-sided Lipschitz condition and polynomial growth conditions. Our idea of proof relies on a generalization of the notion of C-stability, which originated from deterministic numerical analysis of stiff differential equations.
21.10.14
Konstantin Schildknecht (WIAS Berlin)
Simultaneous Statistical Inference for Epigenetic Methylation Data
Abstract: tba
28.10.14
Mario Maurelli (WIAS Berlin)
Mean field equations and a model on lithium-ion batteries
Abstract: In this talk I will introduce interacting particles systems and their mean field limit and I will then discuss a simple model on lithium-ion batteries, where a mean field interpretation arises naturally.
In the first part I will show how a nonlinear partial differential equation (of a certain form) can be interpreted as a mean field limit of an interacting particle system. Advantages of this approach can be both in terms of the physical description of a phenomenon and in terms of simulations.
In the second part I will put this into a particular model arising in the study of lithium-ion batteries. I will then discuss intuition and main challenges of this model.
04.11.14
Roland Hildebrand (WIAS Berlin)
A problem in geometric probability with applications to finance
Abstract: tba
11.11.14
Andreas Andresen (WIAS Berlin)
Two convergence results for an alternation maximizaiton procedure
Abstract: tba
18.11.14
John Schoenmakers (WIAS Berlin)
Regression based duality approach to optimal control, with application to
hydro electricity storage
Abstract: tba
25.11.14
Marcel Ladkau (WIAS Berlin)
Robust optimal stopping
Abstract: tba
02.12.14
Andzhey Koziuk (Moscow Institute of Physics and Technology)
Linear hypothesis
testing for the case with weak instrumental variables
Abstract: In this article novel
approach developed by Vladimir Spokoiny and Mayya Zhilova [6] connecting
real world log-likelihood with bootstrapped one is exploited for the
task of linear hypothesis testing in the problem with weak instrumental
variables [1]. It is assumed that testing hypothesis using such a data
driven approach would provide a better results comparing to usual
asymptotic methods [2],[3] because of asymptotic refinments provided
by bootstrap [4].
[1] Donald W.K. Andrews and Xu Cheng. Estimation and inference with
weak, semi-strong, and strong identification. Econometrica,
80(5):2153-2211, 2012.
[2] Donald W.K. Andrews, Marcelo J. Moreira, and James H.
Stock. Opti- mal two-sided invariant similar tests for instrumental
variables regression. Econometrica, 74(3):715-752, 2006.
[3] Isaiah Andrews. Conditional linear combination tests for weakly
identified models, 2014.
[4] Joel L. Horowitz. The bootstrap, 2000.
[5] Vladimir Spokoiny. Fisher and wilks expansions and
Bernstein - von Mises theorem for growing parameter dimension, 2014.
[6] Vladimir Spokoiny and Mayya Zhilova. Bootstrap confidence
sets under a model misspecification, 2014.
09.12.14
Mikhail Malioutov (Northeastern University)
tba
Abstract: tba
06.01.15
Lars Ruthotto (Emory University Atlanta)
tbc
Abstract: tba
13.01.15
Niklas Willrich (WIAS Berlin)
Results on an automated Lepski's method
Abstract: tba
20.01.15
Moritz Jirak (HU Berlin)
Rate of convergence in the (weighted) CLT under weak dependence
Abstract: tba
03.02.15
Tigran Nagapetyan (WIAS Berlin)
Variance reduced Monte Carlo path simulation method
Abstract: In this talk we present a specially designed control variates for estimating smooth terminal functionals of discretized paths, arising from SDE path approximation. Our control variates decrease the variance of the functional down to the order of discretization step in certain power, which allows us under certain regularity conditions to improve significantly the computational cost / error relation for both Multilevel and Single level Monte Carlo (SMC) path simulation methods. We discuss the additional gains, one can achieve via using weak schemes combined with our control variates for SMC path simulation method. Our results are illustrated by several numerical examples.
10.02.15
Nikolay Baldin (HU Berlin)
A new estimator for the volume of a convex set
Abstract: tba
