A Multi-Agent Algorithm for Financial Derivatives

摘要

Exploiting the parallelism inherent in the lattice approach for option pricing problem, we develop a distributed algorithm for pricing American style options in this study. We have implemented our algorithm on a network of workstations running windows NT. Employing "Aglets" technology, we have assigned parallel tasks to "Mobile Agents" which are distributed to different processors. We describe our agent architecture, distributed algorithm, and study their relative complexities and performance for a class of American-style put options. We observe from the experiments that though the communication overhead due to the network is a significant part of the overall execution time in the performance of the algorithm, "Mobile Agents" provide an excellent environment for parallel computation in a distributed heterogenous network for the class of financial options studied in this paper as well as other complex financial derivatives.