Issue 43, 17 February 2005

In this issue:

For more information about any of the articles featured in this issue, please do not hesitate to contact us at nagnews@nag.co.uk


Top Story - NAG launches the largest commercially available numerical and statistical library


We are delighted to announce the latest version of the NAG Fortran Library, which includes over 250 new functions, taking the total number of user callable routines in Mark 21 to nearly 1500. Unrivaled functionality from the 1467 routines in the globally renowned NAG Library covers the world's most widely used numerical and statistical algorithms.

New functionality includes:

  • New chapter of routines for large scale eigenvalue problems
  • New solvers for large, sparse systems of linear equations
  • Dense linear algebra
    • Many new computational and driver routines from LAPACK 3
    • Linear system black-box routines with error bounds
  • Statistics: extended coverage including
    • Landau and Vavilov distribution and density functions
    • Copulas - Normal and Student's t
    • Multivariate Normal and Student's t distribution random numbers
    • Mixed effects regression
    • Stepwise linear regression
    • Improved quasi-random number generator interfaces
  • Extended optimization capabilities including
    • Nonlinearly constrained minimization for large sparse problems
    • Constrained Quadratic Programming for large sparse problems
  • Additions to the Special Functions chapter

For more information about this new release please email us at mailto:nagnews@nag.co.uk with any questions.


In the News - Getting the Picture, Visualising Financial Data Part 3


In NAGNews 36 and 38 we featured the first two parts of a trilogy of articles by Jeremy Walton, Senior Technical Consultant at NAG, entitled 'Getting the Picture, Visualising Financial Data'. An introduction to the third and final part of this collection of articles, which were published in Financial Engineering News is featured below. Links to parts one and two can be found below.

In the preceding articles in this series, we saw how visualization can help with the understanding of financial data, and looked at some examples using Microsoft' Excel's charting function, along with a more advanced visualization package. While Excel is helpful for plotting one variable as a function of another, such as share price vs. time, we saw that data sets that depend on more than one variable - for example, option volatility - require the use of more sophisticated packages which allow for interactivity when slicing through the data, or selecting three-dimensional viewpoints which are used to display the visualization. In addition, we saw that problems arise in Excel when trying to visualize datasets when the spacing between successive values for variables is irregular, or when the data contains discontinuities.

To read the entire article please visit:

for part one; [site:url]IndustryArticles/VisualizingFinancialData.pdf

for part two; [site:url]IndustryArticles/VisualizingFinancialData–part2.pdf

for part three; [site:url]IndustryArticles/VisualizingFinancialData–part3.pdf


Data Mining - Getting over the "hump"


Decision trees, neural nets, na've Bayesian, genetic algorithms. . . the language of data mining is a brave new world to many of our customers, whether creating applications for internal use or a commercial product. While the potential is great, many are overwhelmed by the choice of technique, interpreting results, and data preparation issues.

In response, NAG has partnered with EWA Systems of Palo Alto, California to complement the software and expertise already available from NAG. EWA brings over ten years of experience in data mining software creation, application building, and results analysis. EWA's great strengths are their complete portfolio of pure Java techniques and tools along with their deep expertise in developing solutions for clients in semiconductor manufacturing, customer relationship management, web analytics/personalization and a variety of other fields.

If you would like to learn more about how data mining techniques can provide the insights needed from your data, contact us at mailto:nagnews@nag.co.uk and we'll assemble the right resources from EWA and NAG to address your challenge.

To read more about NAG's Data Mining & Cleaning Components please visit http://www.nag.co.uk/numeric/DR/drdescription.asp


Tips & Hints - NAG Fortran Library, Mark 21


With some implementations of the NAG Fortran Library, Mark 21, now available, it is appropriate to give some general advice.

Firstly, we recommend sites to upgrade to the new mark whenever possible. This is because a new mark of the Library often includes algorithm improvements and incorporate known bug fixes. An obvious further incentive to upgrade is of course the new routines.

NAG never withdraws a routine without adequate notice. In general at least two marks notice is given of our intention to withdraw a routine. We announce our intentions by placing an asterix beside the routine name in the relevant chapter contents document. On receipt of a new Library it therefore makes sense to check on the status of routines that you expect to be using over a prolonged period of time.

If one of your favourite routines is marked for withdrawal it will have a replacement that we feel is better, so our recommendation is to adapt your program to call this newer routine if at all possible. To help you do this, a document entitled "Advice on Replacement Calls for Withdrawn/Superseded Routines" is available at the front of the manual, in the introductory materials.

If you decide not to do this, then please remember to save the associated documentation for the routine in question. You would be surprised how useful this may be in several years time. An executable compiled and linked with a static library implementation of the NAG Library will continue to work indefinitely. Your problems will only really begin if you subsequently decide to change your program and have not kept the original NAG Library. Shareable libraries are another matter. Because the NAG Library code is not tightly bound into your executables you will need the original library around at all times, if the routine you are calling is no longer in the current library.


Events - DevWeek 2005, London


  • DevWeek, London
    DevWeek is the UK's leading technical conference for software developers, and features expert speakers on a wide range of topics, including Visual Studio 2005 ('Whidbey'), SQL Server 2005 ('Yukon'), C# 2.0, ASP.NET 2.0, Windows Forms 2.0, Security, Debugging and Web Services.

    The 8th annual event takes place in London from 21-25 February 2005, and NAG is delighted to be exhibiting. For more information about this key event in the software development calendar visit the website at www.devweek.co.uk
     
  • CARISMA, London
    *STOP PRESS* - NAG is delighted to be sponsoring the forthcoming CARISMA (The Centre for the Analysis of Risk and Optimisation Modelling Applications) event "New Directions in Risk Modelling and Financial Planning" to be held in London on 18-19 May 2005. For more details click here.


Product News - New Implementations


NAG is committed to offering new implementations of its broad range of numerical and statistical software, compilers and tools. Since the last edition of NAGNews platform availability has increased for the NAG Fortran 90 Library (fl90).

The NAG Fortran 90 Library (fl90), is now also available for the following platform:

  • AMD-64 Linux64 using the NAG f90 Compiler

Managing your subscription


Please feel free to forward this newsletter to colleagues or to post it to your company's Intranet sites.

To subscribe, unsubscribe or learn more about managing your subscription, please see: [site:url]NAGNews/index.asp.