FPGA acceleration of mean variance framework for optimal asset allocation

Abstract

Asset classes respond differently to shifts in financial markets, thus an investor can minimize the risk of loss and maximize return of his portfolio by diversification of assets. Increasing the number of diversified assets in a financial portfolio significantly improves the optimal allocation of different assets giving better investment opportunities. However, a large number of assets require a significant amount of computation that only high performance computing can currently provide. Because of the highly parallel nature of Markowitz' mean variance framework (the most popular approximation approach for optimal asset allocation) an FPGA implementation of the framework can also provide the performance necessary to compute the optimal asset allocation with a large number of assets. In this work, we propose an FPGA implementation of Markowitz' mean variance framework and show it has a potential performance ratio of over a software implementation.

Multi-tenant virtual GPUs for optimising performance of a financial risk application

“…More recently, hardware accelerators with multi-core and many-core processors are employed. For example, financial risk applications are accelerated on Cell BE processors [13,14], FPGAs [15,16] and GPUs [17,18].…”

The hardware accelerator debate: A financial risk case study using many-core computing

“…For example, research related to financial applications exploiting the Cell BE processor is reported in [32,33]. FPGAs are another alternative platform presented in [34,35,36,37]. GPU acceleration is employed more recently [38,39,40,41].…”

Finding the right level of abstraction for minimizing operational expenditure

“…Another option for acceleration is to use Maxeler acceleration cards running dataflow engines on FPGAs (see [22], [23] and [24] for examples of accelerating Monte Carlo with FPGAs, and [25] for other examples of financial computation accelerated with FPGAs). This case study is only a simple example of running a single model repeatedly, while real world applications are likely to be much more complicated, it has been shown in [2] and [22] that FPGAs are appropriate for scaling up to large scale risk computations.…”