Interestingly, it claims to work very well on problems involving large numbers of variables.

Experimental and Agent-Based Computational Economics
http://www1.webng.com/economics/

Amongst other material there ..

1. Global Optimization by Differential Evolution and Particle Swarm Methods: Evaluation on Some Benchmark Functions

2. Construction of Composite Indices -Alternatives to the Indices Obtained by the Principal Components Analysis

3. Computing the Nearest Correlation Matrix from a given Invalid Correlation Matrix

In my defense, let me say that there was no annoyance with the paper nor any attempt to “diss” the algorithm, and my post was directed at the unsatisfactory nature of porting as opposed to rewriting, and an element of frustration in having to wrestle with C.

The code that I was “complaining about” was actually the “Standard PSO 2006″
http://www.particleswarm.info/Standard_PSO_2006.c . As its function is to act as a “reference standard” for researchers, there is certainly a justification for coding it in C.

http://www.particleswarm.info/ says

Quite often some authors say they compare their PSO versions to the “standard one” … which is never the same! So the idea is to define a real standard at least for one year, validated by some researchers of the field, in particular James Kennedy and Maurice Clerc

I did do some research for an appropriate Delphi version and only found one commercial product, and several papers that said that they had implemented the PSO algorithm in Delphi (but no readily accessible code). A quick search for PSO in C# was unproductive. As you say there are matlab versions.

For others reading this post, there is a listing of some of the PSO software here.
http://www.particleswarm.info/Programs.html (note the .info TLD).

Also the PSO homepage is at http://www.cis.syr.edu/~mohan/pso/http://www.cis.syr.edu/~mohan/pso/

If I get the Delphi version going, anyone who is interested can contact me for a copy - but be aware it will not look much like the “standard” , because of the nature of the languages.

There are plenty of people doing current research with particle swarms, coding it in every language from Python to C++ to Java to SAS and Matlab. In a perfect OO world we could grab objects and connect them to our system and they would work; it is likely that you could find such objects on the Internet for particle swarms. Your annoyance here though is unjustified. The paper you linked to was one that introduced the research community to a new algorithm, within weeks of its discovery, and the algorithm has undergone important refinement and revision since then. You should have spent a little effort to search for what you wanted, instead of complaining that the first thing you found didn’t work for you.

Jim Kennedy