Projects

The following research projects are currently performed by our team:

1. Application of Parallel Genetic Algorithms for the Calibration of Financial Models

   • project description:
     The calibration of financial models, e.g. the Cox-Ingersoll-Ross-Model or the Heston-Model, is a very difficult task: In this project we study the possibility to apply parallel genetic algorithms with different migration strategies on different network topologies.
     
     
   • researchers:
     Riccardo Gismondi, Christian Kremnitzer
     
     

2. PVTFR (Parallelized Valutazione Trattamento Fine Rapporto)

   • project description:
     We use the terminology "Valutazione Trattamento Fine Rapporto" since this project was driven by PARAMETRICA, a leading company for financial advisory in Italy and one of the corporate partners of the research institute. The aim of this project is to fulfill a robust estimation of future pension liabilities of any company. This is not just a problem to which banks and insurance companies are confronted. To receive this claimed robust estimation of future pension liabilites, we vary a lot of the input parameters (interest rates, inflation rates, special employee-specific-probabilites, etc.).
     
     
   • researchers:
     Riccardo Gismondi, Christian Kremnitzer
     
     

3. Model Risk and Calibration Risk

   • project description:
     During the last two decades option pricing has witnessed a great variety of new models. Since the restrictions of the standard Black-Scholes model do not correspond to the nowadays observed market prices, one has to take account of different more sophisticated model approaches. This project especially focuses on the estimation and minimization of model- and calibration risk for the Black-Scholes and the Heston Model.
     
     
   • researchers:
     Riccardo Gismondi, Kujtim Avdiu
     
     

4. Just-In-Time Implied Volatility Estimation of Stock Options applying Parallel Computing

   • project description:
     The Black-Scholes-Formula is often used in the backward direction to invert the implied volatility, usually with some solver method. Solvers are slower than closed form approximations, but the latter lack accuracy and often provide option prices exceeding the bid-ask-spreads. This is where the Cluster@WU as our powerfull framework comes in and helps us to solve this problem. For further improvement, the project will also apply the Newton-Raphson iteration technique.
     
     
   • researchers:
     Riccardo Gismondi, Johannes Menzel, Johann Mitlöhner
     
     

5. ALM (Asset Liability Management), Stochastic Optimization and Parallel Computing
   

   • project description:
     ALM is an important part of Enterprise Risk Management (ERM). It can be seen as a framework for assessing and managing risks that yields due to a mismatch between assets and liabilities. The inclusion of stochastic asset modelling turnes ALM into a more powerfull framework. To deal with the high computational requirements originated from this problem, supercomputing will be an absolute necessity.
     
     
   • researchers:
     Riccardo Gismondi, Bernhard Bruckner
     
     

6. Parallel SPES (Standards Pricing Estimator System)

   • project description:
     The platform SPES was developed by PARAMETRICA. It is very powerful and efficient computing framwork for pricing different types of exotic options and structured financial products. To strengthen the performance of SPES, in this project we study the possibility to parallelize SPES applying different paradigms like MPI, OpenMP and MapReduce.
     
     
   • researchers:
     Riccardo Gismondi
     
     

7. Application of Parallel Grammatical Evolution for the Generation of Automated Trading Systems

   • project description:
     Finding automated trading systems for financial derivatives, which yield maximum risk adjusted returns is the ambitious goal of this project. To cope with this difficult task, we use grammatical evolution as proposed by Ryan, Collins, and O'Neill, to evolve potential solutions. By using a certain grammar to map the solutions to actual trading systems, we don't have to worry about the syntactical correctness of the results. In order to improve computation speed and resolving power, multiple generations will be evolved at the same time using parallel computing.
     
     
   • researchers:
     Riccardo Gismondi, Peter Kreuzinger