Skip to Main Content Skip to Search
Accelerating the pace of engineering and science


 

Keynote Presentations

Visualising Financial Data in MATLAB

Tuesday, 19 June, 09:10–09:50

Kerr Hatrick, Deutsche Bank

The visualisation of data has become a vital tool for managing performance and understanding or gaining insight into data, processes, and systems. Two principal reasons for pursuing a range of different representations of financial data are for discovery or insight and for facilitating intuitive or painless comprehension of the data by stakeholders who are less familiar with the detail. In this session, Kerr shows techniques relating to aspects of trading volumes and HFT data analysis to illustrate these interests.


Key Directions for MATLAB

Tuesday, 19 June, 09:50–10:20
Wednesday, 20 June, 09:50–10:20

Roy Lurie, MathWorks

In this session, Roy presents his perspectives on key technologies and trends that are creating both challenges and opportunities in computational finance. He highlights trends in computational resources, quantitative analysis, system integration, and production deployment, and identifies how developments in MATLAB enable quantitative researchers and developers to stay ahead of their peers.


Use of MATLAB for Solvency II Capital Modelling: The Prudential Risk Scenario Generator

Wednesday, 20 June, 09:10–09:50

Sam Bailey, Prudential

In this session, we discuss what the Risk Scenario Generator (RSG) is and how it fits into the Prudential SII capital modelling system. We also discuss the overall architecture of the RSG, including:

  • Key components and interfaces: engines, business modules, user interface, external APIs, and simulation files
  • Plug-and-play separation of business modules from core engines for risk models (calibration and simulation), risk dependency rules (calibration and simulation), simulation validation reports, bootstrapping algorithms, and bootstrapping validation reports

Customer Presentations

"Numerical Methods in Finance" at Imperial College: MATLAB as a Versatile Teaching Tool

Tuesday, 19 June, 10:50–11:30

Antoine Jacquier, Imperial College London

Imperial College’s M.Sc. in mathematics and finance prepares students for the challenges of working in finance, combining education in key mathematical techniques with practical aspects of working with industrial partners, including investment banks and hedge funds. Students learn probability, stochastic analysis, optimisation, numerical methods, and other techniques through student internships, engaging in joint research consultancy arrangements, and other forms of collaboration. In this session, Antoine talks about the role of MATLAB in the syllabus, particularly for course projects where students demonstrate their understanding of key mathematical techniques such as finite difference methods, through their application in a reasonably structured, yet not excessively complex, computational finance environment. This approach enables focus on the algorithm, but not at the expense of good programming.


Using MATLAB for Risk Modelling: Two Practical Applications

Tuesday, 19 June, 13:30–14:10

Gary Dunn and Evi Pliota, HSBC

In this session, Gary and Evi present two applications of using MATLAB for risk modelling: Incremental Risk Charge (IRC) and HSBC's De-peg Risk Measure (DPRM).

IRC is a regulatory capital model required to capture default and credit migration risk in the trading book. HSBC has had an approved IRC model since 2008; however, some enhancements were required for Basel III rules for the trading book that came into force at the end of 2011. MATLAB was used to build a replica of the model in production, which was then used to research the necessary enhancements. The MATLAB model is now often used for analysis of production results and for what-if analysis. The presentation discusses the IRC model requirements, MATLAB implementation, and experimentation with GPU technology to enhance performance.

The DPRM model calculates the capital requirement for the risk of the peg to be abolished or the regime to change. For certain currencies (pegged or heavily managed), the spot exchange rate is pegged at a fixed rate (typically to USD) or managed within a predefined band around a pegged rate. Historic FX rate scenarios for pegged or managed currencies typically display low volatility; therefore, a VaR measure calculated using historical movements will be understated as it does not reflect the risk of a peg break and change in currency regime. The purpose of the DPRM described in this presentation is to capture the risk of peg break and generate the appropriate capital requirement through the capital add-on calculation. MATLAB is used to build the model, and the application is shared with the Market Risk Control and Market Risk Managers using MATLAB Compiler.


Using MATLAB for Real-Measure Calibration of Stochastic Volatility Models in Finance

Tuesday, 19 June, 15:25–16:05

Leonid Timochouk, Royal Bank of Scotland, and Vadim Anufrijenko, independent trader

In many financial applications (such as volatility arbitrage trading, options market making, algorithmic trading strategies, counter-party credit exposure computation, VaR analysis, and others), it is important to construct probability density functions (PDFs) of the underlying stochastic processes in real measure. In other words, the parameters of the corresponding stochastic/local volatility (SLV) models are to be calibrated to the time series of the observable price spot/futures price values rather than market prices of options. One way of performing such a calibration is by applying Bayesian optimal filtering with conditioning on price observations. This method requires computation of transition probabilities between conditioning points. In this session, we present two solutions for the latter problem, both implemented in MATLAB. One solution uses a generalised Fokker-Planck PDE, and the other is based on a semi-analytical method of heat kernel expansions. The pros and cons of both solutions are discussed, as well as lessons learned about using MATLAB for this type of problem.


Risk Management of a Composite Commodity Portfolio Using Monte Carlo Simulation and MATLAB

Wednesday, 20 June, 10:50–11:30

Gianluca Fusai, Cass Business School

In this session, we discuss how to properly assess the risk-return tradeoff of a composite commodity portfolio. We examine estimation versus calibration issues and look at a real-world case study of a complex commodity portfolio, which will be presented via Monte Carlo simulation. Risk contribution of the portfolio components and how to find the best hedge are also discussed, through a MATLAB implementation.


Client-Facing Products with MATLAB

Wednesday, 20 June, 13:30–14:10

Chris Squirrell and Tim Lock, Aon Hewitt

Aon Hewitt works with a broad spectrum of clients, helping them to identify, quantify, and manage financial risks. In this session, we focus on the use of MATLAB and other technologies in a tool developed specifically for the client-facing consulting environment. PRisM (Pension Risk Modelling) is able to clearly illustrate and articulate the consequences of key investment and risk management decisions versus our clients' objectives.


Measuring and Monitoring Market Risk Using MATLAB

Wednesday, 20 June, 15:25–16:05

Athanasios Bolmatis, Fulcrum Asset Management

With an experienced investment team of economists, portfolio managers, and risk experts, Fulcrum Asset Management invests across global markets and asset classes with a range of absolute and relative return strategies. Their proprietary risk management techniques are designed to maximise risk adjusted returns. Fulcrum regards risk management as an integral part of portfolio management. This session shows how Fulcrum measures risk using MATLAB and illustrates, using actual data, the importance of risk management in improving the risk-reward profile of a strategy.

MathWorks Presentations

MATLAB: Committed to Excellent Algorithms

Tuesday, 19 June, 11:30–12:15
Wednesday, 20 June, 11:30–12:15

Jos Martin, MathWorks

In this session, Jos discusses the science applied to develop and improve the core algorithms in MATLAB. Jos focuses on the opportunities—and challenges—in ensuring MATLAB continues to be the tool of choice in the world of multicore CPUs, in grids and clouds, on mobile platforms, and on GPUs. He discusses the processes involved in developing and testing MATLAB to drive usability features, development environment enhancements, and overall reliability.


Mathematics, Optimisation, and Statistics: Cutting-Edge Algorithms for Difficult Data and Complex Models

Tuesday, 19 June, 14:10–14:55
Wednesday, 20 June, 14:10–14:55

Ian Noell, MathWorks

In this session, Ian details key enhancements in numerical modelling in MATLAB. The session highlights several optimisation problems and describes the plethora of algorithms, many new, available to solve them. Ian advises on architecting and programming optimisation models and how to use MATLAB to interpret and test solutions. He then outlines statistical developments relevant to high-dimensional and large data sets, examining regression, machine learning, and feature selection algorithms, such as discriminant analysis, loess regression, bagged decision trees, support vector machines, and neural networks. Ian also summarises new capabilities in symbolic math and curve fitting.


Taking MATLAB into Production

Tuesday, 19 June, 16:05–16:45
Wednesday, 20 June, 16:05–16:45

David Sampson, MathWorks

Financial organisations are increasingly relying on MATLAB in mission-critical production software environments in trading, risk, and insurance. In this session, David explores issues of collaboration, robustness, and software integration that are particularly important when developing for such environments. First, he describes how to architect your MATLAB applications to allow software engineers and quantitative analysts to work together efficiently. Second, he discusses implementation and testing practices to improve the robustness of your MATLAB applications. Finally, he explores how to integrate your MATLAB components into external code for deployment, plus how to integrate your external code into MATLAB for development.

Master Classes

Optimisation and Curve Fitting

Tuesday, 19 June, 11:30–12:15
Wednesday, 20 June, 11:30–12:15

Ian Noell, MathWorks

This master class presents two examples. In the first example, Ian outlines financial and econometric problems that can benefit from good optimisation. He highlights effective optimisation solvers and addresses common challenges users face when applying them. Ian details advanced options to test, improve, and speed up solutions. In the second example, he shows the curve fitting tools available in MATLAB, with particular reference to yield curves, including spline techniques and supplying custom equations such as Nelson Siegel. Ian concludes by showing a practical yield curve framework, allowing for smoothness when you have long yields, for example.


MATLAB Object-Oriented Programming

Tuesday, 19 June, 14:10–14:55
Wednesday, 20 June, 14:10–14:55

David Sampson, MathWorks

As individuals and groups create larger and more sophisticated applications in MATLAB, they must manage the additional complexity that comes with more code and more developers, while continuing to provide uncomplicated programming interfaces for analysts and occasional users. In this master class, David discusses some useful application architectures that can be used in MATLAB and highlights a number of relevant language features of the MATLAB object system. He includes examples from trading, risk, and insurance.


MATLAB: Parallel Computing and GPUs

Tuesday, 19 June, 16:05–16:45
Wednesday, 20 June, 16:05–16:45

Ben Tordoff, MathWorks

Most of us have multiple cores as well as a graphics processing unit (GPU) in our desktop computer. We can also access grids and clouds, all of which have the potential to speed up numerical computing. In this master class, Ben looks at the sorts of calculations that can be sped up by parallel computing, using multiple cores on one machine, a cluster of machines, or a GPU. Detailed examples are used to examine potential problems and solutions when converting algorithms for parallel execution.