Computational Finance is an interdisciplinary branch of scientific computing. The main goal is to determine the risks generated by financial products as accurately as possible. Application areas are the evaluation and the trading of securities, stocks, and bonds and the determination of sensitivities and hedging strategies, risk assessment, asset-liability management, as well as investment decisions and strategic corporate planning. Current challenges are increasingly complex financial products, market models with multiple sources of uncertainty and the simultaneous management of assets and liabilities as optimization problems.
SCAI‘s business area Computational Finance develops efficient and robust numerical algorithms and implements them on parallel high-performance computing systems. Current areas of research are:
- Dimension-adaptive sparse grid quadrature
- Multilevel quasi-Monte-Carlo simulation
- Machine learning: deep learning, tensor methods, generalized support vector machines, transfer learning, reinforcement learning, nonlinear dimensionality reduction
These techniques allow calculations with high accuracy while substantially reducing computation times. The aim of the department is to develop a faster and qualitatively better analysis of financial data for predictions and, consequently, for well-founded investment decisions. Typical applications for machine learning methods are, for example, investment predictions, risk management, and algorithmic trading. We describe three examples in detail.