Lunch Seminar Economics (LSE) - 2020/2021
Interviene: Georg Pflug, Universität Wien
Link: aula virtuale Teams
We review some facts about approximation of general stochastic processes by simpler ones, especially valuated trees. This problem is relevant for dynamic stochastic programming, where trees are the important data structure. Since trees can be defined as a recursive structure, we emphasize the recursive character of some algorithms. Several older results will be put into perspective and some new results about fast calculation of distances between stochastic processes as well as about approximation of a larger tree by a simpler one - known as scenario reduction - will be presented. The presented algorithms (in Julia) are public and are collected in a GitHub library, where also the documentation can be found.