In this issue:
Latest NAG Library features new solvers in the Optimization Modelling Suite
The NAG Library brings substantial benefits to quants and other finance professionals because of its exceptional quality, accuracy and flexibility. It eliminates the lengthy time of writing your own code and the uncertainty that comes with using open source codes. Over 70% of Tier 1 banks use the NAG Library alongside their existing quant libraries to ensure their calculations give fast and correct answers.
A new release of our flagship product, the NAG Library, is available for download. Mark 26 brings a multitude of new functionality including new Optimization solvers in the NAG Optimization Modelling Suite. The new Library also features additional routines in existing areas of Nearest Correlation Matrix, Quadrature, Least Squares and Eigenvalue Problems (LAPACK) and OpenMP Utilities.
The NAG Optimization Modelling Suite has been introduced to better tackle the input of complex problems without forming difficult interfaces with a daunting number of arguments. It is available for the new optimization solvers introduced at this mark, the semidefinite programming solver and the interior point method for nonlinear optimization. However, the suite will expand in the years to come for more problem types.
Mark 26 new functionality highlights (click on the functionality to learn more):
- Interior Point Method for large-scale nonlinear programming problems (accessible from the new
Optimization Modelling Suite)
- Linear and nonlinear semidefinite programming solver
- Gaussian Quadrature
- Nearest Correlation Matrix functions
- Least Squares and Eigenvalue Problems (LAPACK)
- OpenMP Utilities
How can I start using Mark 26?
You might have existing NAG Library entitlement as part of a NAG Software Agreement at your organisation. If you currently use the NAG Library and would like us to see if you're eligible for an upgrade to the new mark, email us and we'll do the checking. If you're interested in using the routines in the Library do get in touch or visit our website for more information.
Technical Report: Index-tracking Portfolio Optimization Model
Index-tracking is a form of passive fund management. The index-tracking problem is the problem of reproducing the performance of a stock market index by considering a portfolio of assets comprised on the index. This approach differs from the full replication strategy, where a fund purchases all the stocks that make up a particular market index. A passively managed fund whose objective is to reproduce the return on an index is known as an index fund, or a tracker fund. The classical index-tracking approach represents the problem in a least square framework with errors computed using a sample of historical data. A new approach described in N. C. P. Edirisinghe "Index-tracking optimal portfolio selection" Quantitative Finance Letters, Vol. 1, 16-20, (2013), which this report is based on, replaces the classical risk measure of portfolio variance by the variance of tracking errors between stochastic index return and the return on the portfolio selection. In this report we formulate the index-tracking portfolio optimization model and present an illustrative example where we compare the presented model with the classical Markowitz mean-variance portfolio optimization model.
You can read the report here.
Investment Company Utilize NAG Optimization Solvers to Calibrate Nonlinear Least Squares Problem
Global investment company Exane specialize in three finance areas: Cash Equities, Derivatives and Asset Management, and it was within the Equity Derivatives function that Exane benefitted from using NAG's superior optimization solvers to effectively calibrate parametric arbitrage-free volatility surfaces.
The Exane Quant Team for Equity Derivatives needed to quickly, efficiently and, on a continuous basis, solve a constrained nonlinear least squares optimization problem with approximately 50 parameters, 100 linear constraints and 100 nonlinear constraints.
They selected NAG Library optimization routines and conducted an extensive test phase, including pitching them against other numerical libraries and several open source routines. During the testing phase NAG experts helped the Exane team achieve a proof of concept, overcoming the initial complexity challenges. At the end of the evaluation NAG was chosen to supply their solvers for a host of reasons including the extensive algorithmic coverage found in the NAG Library. The Library offers numerous algorithms for the same class of problems which means the user can choose exactly the right solver for the problem.
How to Calculate a Nearest Correlation Matrix webinar recording and learning material
NAG recently presented a Nearest Correlation Matrix webinar. This 30-minute webinar presented by Nearest Correlation Matrix expert Dr Craig Lucas teaches how to deal with the issues in forming a nearest correlation matrix from real data including:
- How issues with data can lead to approximate correlation matrices
- Cover some theoretical approaches and,
- Explain how to use a set of alternative specialized routines to compute true correlation matrices, while fixing some of the original entries
- Features practical examples using Python
View the webinar on YouTube.
Read the learning material on GitHub.
Technical CVA Research Report plus London Event - Coming in 2017
NAG is currently working on a CVA research project which will result in a technical report in the new year. The work will demonstrate how NAG combines their Algorithmic Differentiation (AD) tool dco/c++ with the latest C++11 features and GPUs to deliver incredible performance on CVA code. This research will be presented during Spring 2017. You'll receive an invitation to this event in due course.
NAG Continues to Pioneer the AAD Revolution in Quant Finance
Adoption of adjoint methods in quant codes has been growing in popularity in recent years, thanks to evangelists such as Giles and Glasserman - Smoking Adjoints: fast Monte Carlo Greeks, Capriotti - Fast Greeks by algorithmic differentiation as well as Naumann and Du Toit - Adjoint Algorithmic Differentiation Tool Support for Typical Numerical Patterns in Computational Finance.
NAG helps clients implement adjoint methods by providing
- Algorithmic Differentiation (AD) Tools (dco/c++)
- Adjoint implementations of numerical kernels (for CPU and accelerator such as NVIDIA GPU)
- Adjoint implementations of client codes
Building on this considerable experience, in NAG's next release of the NAG Library we will provide adjoint versions of several routines. If you would like to ensure an adjoint version of your favourite NAG routine is included in our first release please contact us direct.
Mathematical Optimization Consultancy - enabling fast and accurate decision making
Mathematical Optimization, also known as Mathematical Programming, is an aid for decision making utilized on a grand scale across all industries. Advanced analytical techniques are used to find the best value of the inputs from a given set which is specified by physical limits of the problem and user's restrictions. The quality of the result is measured by a user metric provided as a scalar function of the inputs. Optimization problems come from a massively diverse range of fields and industries, such as portfolio optimization or calibration in finance, structural optimization in engineering, data fitting in weather forecasting, parameter estimation in chemistry and many more.
NAG's Mathematical Optimization Consultancy team provide you with all the information needed to solve an optimization problem. They will discuss your current approach and help guide you to use the right solver for the particular problem. If you need to improve the performance, they can assess how your current model fits the particular solver and advise on a possible reformulation or tune the solver for you. If the problem requires it, NAG will develop an entirely new solver or adapt an existing solver.
To reinforce our commitment to NAG's Mathematical Optimization Consultancy we will provide free 1:1 consulting sessions with NAG consultants to readers of this newsletter. Book your time with our consultants by quoting QFNews. Share your headaches with us, send us your models, questions or simply book your call by replying to this newsletter or writing to email@example.com quoting QFNews. Appointments subject to availability, typical consultations last an hour and are capped at two hours.
Best of the Blog
- Improved Accessibility for NAG routines for Python Data Scientists
- Portfolio Credit Risk
- Changepoint Analysis using MongoDB and the NAG Library for Python