C(omp)++ | A Portal for the Computational Finance Community

Is Computational Finance a Multi-Disciplinary Activity? by Daniel Duffy

Software is an indispensable part of most businesses. In the case of computational finance the main challenge is to propose new financial models, define mathematical structures and algorithms for them and then map these algorithms to code. Some of the questions that arise are:

• How do we define a repeatable process for modelling derivatives in C++ or C#?

• How do the financial, mathematical and software models link up?

• What are the requirements for our resulting software products in terms of maintainability, efficiency, functionality and reliability?

These questions are more easily posed than answered for a number of reasons; there are a number of challenges to be resolved. First, the ideal development team consists of quant analysts, developers and other professionals with knowledge of finance, mathematics and numerical analysis. Second, the members of the team need to have a common language (or even several languages) that they can use to promote interoperability. Finally, we need to take account of constraints such as time-to-market, project schedule and budget as well as the requirements on the resulting software product.
The techniques that we can use to promote computational finance as a multidisciplinary activity are:

• Good mathematical and numerical models that unambiguously define derivatives products

• Using the de facto modeling language UML (Unified Modeling Language)

• Using proven Design and System Patterns

• Using multi-paradigm programming languages that support the object-oriented and generic programming models

I shall be speaking about these software-related issues at the free evening seminar in the City of London on June 26 2007 at Charing Cross. More information and registration (you must register in order to attend), see http://www.datasimfinancial.com/course_detail.php?courseId=13

For queries, please contact Ilona Hooft Graafland at ilona@datasim.nl

Let me know if you have any questions.

Daniel

 

Luigi Ballabio: A first look into QuantLib

If you happened to have a look at QuantLib, it won't have escaped you that the library sorely lacks documentation. I'm not talking of some kind of class reference; that one is available---mostly because it can be generated automatically. What is missing is a description of the architecture of the library, and of how the various classes interact.

Well, as the song says, rejoice, all ye faithful: such documentation is finally being written. Here is a draft of a chapter. You're welcome to read it; I'll be grateful for any feedback, corrections, and criticisms. In particular, criticisms will help us improve both the docs and the library itself, so fire ahead.

Luigi

Email me!

Download Chapter 2 here

(right-click & save target to download this file)

Luigi Ballabio: Show Me the Code!

As I was checking the new entries in my RSS feeds, I recently stumbled upon this blog entry. Go ahead, read it. I'll wait right here.

Are you back, gentle reader? What do you think? Although the blogger referred in particular to computer science, his argument applies to almost all fields. In fact, Mark Joshi made the very same point in a conversation we had last summer. How much can you trust the results of a quantitative-finance paper if they're not repeatable?

Of course, one could reproduce the results - in a way. Based on the author's description, one could code the same calculations. Like we have that kind of time, right?

In fact, the insight toolkit mentioned at the end of the post is a nice idea. If authors wrote their code on top of such a library and made it available, the advantage would be twofold. On the one hand, they would build upon classes and functions reviewed and tested by scores of other users. On the other hand, readers familiar with the toolkit would find it much easier to read the author's code and could focus on understanding the actual new algorithms rather than the scaffolding.

Need I mention that we have a similar toolkit for quantitative finance?

Thoughts anyone? I'll be glad to read them.

Luigi