QuantDay 2007 Quantitative Modeling & Financial Market Dynamics October 31, 2007
The event was held in the heart of New York City's Financial District from 4:00 to 8:00 PM at New York Marriott Financial Center, 85 West Street (website, googlemap), just a few steps from Wall Street and around the corner from the American Stock Exchange.
In addition to the program of world-class speakers (as outlined below), this event featured a Technical Q&A pre-event from 2:30 to 3:30 PM and a cocktail reception following the last presentation from 8:00 to 9:30 PM.
2:30 - 3:30
Technical Q&A pre-event: NAG personnel was on hand to answer technical questions and demonstrate linking to the NAG Library from different environments such as Maple, MATLAB, C/C++ -and others.
3:30 - 4:00
Registration and Check In: QuantDay 2007
4:00 - 4:30
NAG Recognition of Wilmott Magazine with remarks by Dr. Paul Wilmott
4:30 - 5:30
Numerical Software, Market Data and Extreme Events
(click here to download the powerpoint slides [3.88MB] for this talk)
Dr. Robert Tong, Numerical Algorithms Group, Ltd.
Abstract: Financial markets produce large quantities of high-frequency data which must be processed and analyzed in order to inform trading decisions. Software systems are therefore essential to competitiveness in this environment. The steps leading from the pre-processing of tick data through to analysis of underlying patterns throw up many interesting problems for the design and implementation of numerical software. These can be viewed as independent components for functions such as data cleaning and wavelet analysis, each with their theoretical underpinnings. However, the response of the software to extreme events reflected in market data must be considered, both in terms of the individual numerical components and as part of a more complex software system. Aspects of numerical software development in this context are discussed with reference to FX trading and in terms of specific techniques, such as wavelet analysis.5:30 - 5:45
5:45 - 6:45
Multilevel Monte Carlo Path Simulation
(click here to download the pdf pages [348KB] for this talk)
Dr. Mike Giles, Oxford University Computing Laboratory
Abstract: This presentation will address the multilevel method which combines results from Monte Carlo path simulations with differing time steps to achieve a user-specified accuracy at a much lower cost than standard Monte Carlo. The ideas will be introduced with a simple Euler discretization of a European call option, and then extended to the higher-order Milstein method and Asian, lookback, barrier and digital options. Further computational savings will be demonstrated by combining the technique with Quasi Monte Carlo methods.6:45 - 7:00
7:00 - 8:00
Dynamic Portfolio Optimization using Decomposition and Finite-Element Methods
(click here to download the powerpoint slides [255KB] for this talk)
Dr. John Birge, Co-Founder of Quantstar, University of Chicago GSB
Abstract: In dynamic environments, asset allocations based on static portfolio models can lead to excessive transaction costs or significant imbalances. Dynamic portfolio allocation can provide improved performance by balancing transaction costs with efficient allocation and response to market conditions and benchmark positions. The resulting models can quickly become quite large and complex, requiring some form of approximation for solution. This talk will describe the range of potential approximation methods and in particular will focus on serial decomposition using global value function outer approximation and on solution of the corresponding variational inequality using a finite-element approach with high-order basis functions.8:00 - 9:30
NAG has a rich, 30-year history of collaboration in a variety of sectors, one of which is Finance. NAG software components underpin the models and applications for many of the world's most prominent financial institutions ' and this event is a natural extension of our ongoing collaborative relationships within this sector.