Re: [Boost-users] Queries of a Boost newbie

From: Johan Råde (rade_at_[hidden])
Date: 2008-05-02 15:19:05

ajay gopalakrishnan wrote:

> I work mainly in Machine Learning & Data Mining and this job mainly 
> requires me to write very efficient and fast code for numerical 
> processing programs as well as do a lot of data, especially text 
> preprocessing. It would be great if you cant point to me some good 
> packages in Boost for the following tasks. Packages that are very good 
> and don't have a very steep learning curve.
> 
>     * Text Parsing. package that lets me do something like what Sed and
>       Awk can do.

Take a look at Boost.Regex and Boost.Spirit.

>     * Linear Algebra - Eigen analysis, matrix operations, Matrix
>       decomposition etc.

There is Boost.uBLAS. But frankly, I don't like uBLAS.
I use my own generic C++ wrapper around Intel MKL and VecLib.

>     * Optimization routines - Linear programming , Quadratic Programming etc.
>     * HTML, XML parsing etc.

Boost does not have much to offer in these areas.
Unfortunately most Boost libraries are written by a single person.
The Boost culture does not seem to encourage collaborations.
And these areas would require teams of several developers to get something useful done.

----
Also, do take a look at smart pointers, bind, function, filesystem, format and thread.
They solve a lot of general software development problems.
You mentioned data mining. If you do statistical analysis of data,
then you will need Boost.Math.Statistical Distributions.
HTH,
Johan Råde