Abstract

Portfolios of financial instruments are becoming larger, more diversified, and increasingly contain more complex instruments. Therefore, the modeling of the behavior of such portfolios yields increasingly demanding computational problems. In addition there is demand for increased speed to solve these computational portfolio problems (such as evaluation, portfolio rebalancing and optimization, and value-at-risk computations). The demand for more speed is driven, in part, by the move toward automated and semi-automated trading systems.

In this talk we discuss the two basic approaches to increasing speed in portfolio computations, algorithmic cleverness and increased compute power via use of more processors, and discuss when to apply one or the other. We give some examples.

Biography of Speaker

Professor Coleman has wide experience with US government research laboratories doing fundamental algorithmic research, and the financial services industry working with companies in New York, Toronto, Hong Kong, Tokyo, and Singapore. Professor Coleman founded, and for eight years directed, an industry research centre in Manhattan working with Wall Street firms on quantitative finance and risk management problems. Professor Coleman now co-directs the Cornell-Waterloo Solutions Lab in Manhattan, a centre devoted to facilitating academic/industry research work in the financial computing industry.

Professor Coleman’s research interests involve the design and implementation of efficient practical algorithms for large-scale optimization problems and their applications. Efficiency concerns, in turn, lead to research into the effective use of parallelism, the application of automatic differentiation, the exploitation of sparsity and structure in the problem formulation, and effective ways to deal with constraints. In recent years he has been most concerned with computational finance applications: examples include portfolio optimization, liquidation, optimal hedging approaches (in incomplete markets), and volatility surface calculations.