Stránka o brožuře věnované vybraným softwarům pro řešení úloh stochastického programování ZDE.

Program bude průběžně doplňován. Hosté jsou srdečně zváni.

• Seminář se nekoná

  Autor:

  Datum:

  22.2.2018

• Úvodní přednáška

  Autor:

  doc. RNDr. Ing. Miloš Kopa, Ph.D.

  Datum:

  1.3.2018

• Portfolio Optimization with DARA Stochastic Dominance Constraints

  Autor:

  doc. RNDr. Ing. Miloš Kopa, Ph.D.

  Datum:

  8.3.2018

• Multistage multivariate nested distance: stage and scenario reduction

  Autor:

  Sebastiano Vitali, Ph.D.

  Datum:

  15.3.2018

  Abstract:

  Multistage stochastic optimization requires the definition and the generation of a discrete stochastic tree that represents the evolution of the uncertain parameters through the time and the space. The dimension of the tree is the results of a trade-off between adaptability to the original probability distribution and computational tractability. Therefore, the recent literature investigates the concept of distance between trees which are candidate to be adopted as stochastic framework for the multistage model optimization. The contribution of this paper is to compute the nested distance between a large set of multistage and multivariate trees and, for a sample of basics financial problem, to empirically show the positive relation between the tree distance and the distance between the corresponding optimal solutions and the optimal objective values. Moreover, analyse the loss (measured in term of nested distance) due to a scenario reduction or to a stage reduction.

• MS++: Library for Risk-averse Multistage Stochastic Programming

  Autor:

  RNDr. Martin Šmíd, Ph.D.

  Datum:

  22.3.2018

• Seminář se nekoná - Velikonoce

  Autor:

  Datum:

  29.3.2018

• Guaranteed Bounds for Multistage Risk-Averse Stochastic Optimization Programs

  Autor:

  Assoc. Prof. Francesca Maggioni

  Datum:

  5.4.2018

  Abstract:

  In general, multistage stochastic optimization problems are formulated on the basis of continuous distributions describing the uncertainty. Such "infinite" problems are practically impossible to solve as they are formulated and finite tree approximations of the underlying stochastic processes are used as proxies. In this talk bounding methods for multistage stochastic optimization problems are discussed. First we consider bounds based on the assumption that a sufficiently large discretized scenario tree describing the problem uncertainty is given but is unsolvable. Monotonic bounds based on group subproblems of the large scenario tree will be discussed and compared in terms of computational complexity. Secondly, we demonstrate how one can find guaranteed bounds, i.e. finite tree models, for which the optimal values give upper and lower bounds for the optimal values of the original infinite problem. We consider approximations in the first order stochastic sense, in the convex order sense and based on subgradient approximations. Their use is shown in a multistage risk-averse production problem.
  Work done in collaboration with Prof. Georg Pflug (University of Vienna).

• tbs

  Autor:

  RNDr. Michal Houda, Ph.D.

  Datum:

  12.4.2018

• Distributionally robust chance-constrained dynamic pension fund management

  Autor:

  Prof. Giorgio Consigli

  Datum:

  19.4.2018

  Abstract:

  We consider a canonical asset-liability management (ALM) model for a defined benefit pension fund from the perspective of a PF manager seeking an optimal dynamic investment strategy under a set of asset and liability constraints and in particular a chance constraint on the pension fund solvency condition. This class of problem is well-known and it has been studied under several modeling approaches, and specifically within a discrete framework through multistage stochastic programming (MSP). A real-world case-study has been presented with a detailed problem formulation and MSP solution approach in {Consigli et al. 2017: as in what follows, the complexity of such problem class comes from its long-term nature and the underlying risk sources, affecting asset returns and liability flows. In a MSP framwork those uncertainties require a dedicated statistical model from which a scenario tree process is derived. When, as mostly the case, asset returns and liability costs are assumed to carry a continuous probability space, approximate solutions can be obtained by substituting those probability distribution with a discrete approximation and allowing strategy revisions only at discrete time points. In presence of realistic PF ALM problems' instances MSP approaches are able to accommodate a rich set of assumptions and market details but at the cost of a possible curse of dimensionalty, the problem's in-sample instability and significant model risk: the first two represent a non trivial trade-off, since in-sample stability calls for robust and stable solutions to different, sufficiently rich sampling methods. The latter may lead to inefficient decision processes due to unsuitable statistical assumptions. These drawbacks may all be overcome through a distributonally robust optimization (DRO) approach, that can be regarded as a natural generalization of stochastic programming and robust optimization approaches, accounting for both the decision maker's attitude to risk and ambiguity: the latter being refered to the uncertainty characterizing the probability measure to be associated with the decision problem's underlying stochastic factors.

• tbs

  Autor:

  RNDr. Mgr. Barbora Petrová

  Datum:

  26.4.2018

• Optimal Loan Performance Management via Stochasic Programming

  Autor:

  Mgr. Tomáš Rusý

  Datum:

  3.5.2018

• tbs

  Autor:

  RNDr. Vlasta Kaňková, CSc.

  Datum:

  10.5.2018

• Seminář se nekoná - Konference Kaunas

  Autor:

  Datum:

  17.5.2018

• Seminář se nekoná

  Autor:

  Datum:

  24.5.2018