Program bude průběžně doplňován. Hosté jsou srdečně zváni. Zoom meeting

• Asset pricing and interest rates under extreme climate change financial risks

  Autor:

  Davide Radi

  Datum:

  27.02.2025

• Úvodní seminář

  Autor:

  Miloš Kopa

  Datum:

  06.03.2025

• Seminář se nekoná

  Autor:

  Seminář se nekoná

  Datum:

  13.03.2025

• Optimality and Fairness: application to reimbursement of COVID-19 vaccines

  Autor:

  Takaki Hayashi

  Datum:

  20.03.2025

  Abstract:

  We intend to explore low latency, lead-lag relationships between stock prices using tick data. A salient feature of tick data is its irregularity of intervals between observation times, stemming from records being timestamped at the exact moments when trades or orders occur. A natural way to handle with is to model tick data as discrete observations from continuous-time stochastic processes. A pioneering work is done by Hoffmann, Rosenbaum and Yoshida (2013), who propose a method to estimate the unknown lead-lag times based cross-correlations in a semimartigale framework. Alternatively, Dobrev and Schaumburg (2016) propose a method based on counts of co-arrival times of point processes, which can capture low latency, cross-market activities by high-freqeuncy traders. 　In this talk, we briefly review these approaches, followed by our stochastic model and wavelet-based estimation method for lead-lag times. It is specifically designed to capture multiple lead-lag relationships that may coexist within two time series. 　We present two empirical analyses conducted using those approaches. First, we investigate lead-lag relationships between quotes on the NASDAQ and BATS exchanges for each stock of the NASDAQ 100 Index. Second, we investigate lead-lag relationships between quotes of Nikkei 225 index-linked ETFs on the Tokyo Stock Exchange (TSE) and examine whether/how the ETF market-making system introduced by TSE in 2018 and 2019, as well as the system upgrade of its *arrowhead* (the trading system for cash products) implemented in November 2019, have influenced lead-lag relationships among those ETFs. Our approach is shown to have capacity to capture complex nature of these lead-lag relationships in a signle model.

• Uncertainty in real-world vehicle routing

  Autor:

  Václav Sobotka

  Datum:

  27.03.2025

  Abstrakt:

  The motivation for our research arises from the limitations of traditional deterministic heuristic solvers for vehicle routing problems (VRP) observed in industrial practice. In general, the quantities provided to the solvers as inputs, e.g., loads or service times, are typically estimates or simplifying reflections of reality. Moreover, the input information may be incomplete as new customers may request service while the vehicles are already on the road. Current state-of-the-art solvers are applicable to complex VRP variants at scale. However, their inability to reason about such uncertainties limits their usefulness in real-world applications. Despite stochastic VRPs being a widely studied topic, related approaches are typically centered around the uncertainty in the problem rather than extending successful and generic deterministic methods. Moreover, uncertainty-related methodologies and models are often strongly linked to computationally expensive sampling or exact algorithms making their scaling problematic. In this talk, we will discuss how to naturally extend the state-of-the-art methods to account for uncertainties without sacrificing their crucial qualities - generality and scalability.

• TBA

  Autor:

  Monika Kalatová

  Datum:

  03.04.2025

• TBA

  Autor:

  Jana Junová

  Datum:

  10.04.2025

• Seminář se nekoná

  Autor:

  Seminář se nekoná

  Datum:

  17.04.2025

• Seminář zrušen

  Autor:

  Seminář zrušen

  Datum:

  24.04.2025

• Seminář se nekoná

  Autor:

  Seminář se nekoná

  Datum:

  01.05.2025

• Seminář se nekoná

  Autor:

  Seminář se nekoná

  Datum:

  08.05.2025

• Problem-driven scenario generation for optimization under uncertainty

  Autor:

  Jamie Fairbrother

  Datum:

  15.05.2025

  Abstrakt:

  In optimization problems with uncertainty, it is common to represent uncertain parameters by a finite set of potential future scenarios. For example, in portfolio selection problems, each scenario might correspond to a set possible of returns for a collection of financial assets. Sets of scenarios can be constructed in many ways. For example, they may come directly from historical data or be modelled with some (discretized) probability distribution. Wherever the scenarios come from, there is a usually a fundamental trade-off: the more scenarios one uses, the more detailed the representation of the uncertainty, but on the other hand, the less tractable the resulting optimization problem. Scenario generation concerns the construction of a set of scenarios for use in an optimization problem, and the aim is typically to represent uncertainty in a concise a manner as possible. Traditionally scenario generation approaches have been input or distribution-driven. That is, they construct scenarios in such a way to match some reference set or probability distribution without explicitly considering the underlying optimization. More recently problem-driven methods to scenario generation have been proposed which exploit problem structure to provide a more concise representation of uncertainty. In this talk we present some recent problem-driven approaches to scenario generation, concentrating in particular on approaches for two-stage stochastic programming.

• TBA

  Autor:

  Monika Matoušková

  Datum:

  22.05.2025