05:22 < ANUPAM> HELLO EVERYONE!! I AM ANUPAM TRIPATHI (UNDERGRAD AT IIT-BHU). I WANT TO JOIN MLPACK IN GSOC PROGRAM THIS SUMMER.

07:25 < rcurtin> ANUPAM: welcome! I think your caps lock key might be broken :) anyway, you might be interested in reading http://www.mlpack.org/gsoc.html for more information about GSoC

09:39 < suryo> Hello zoq. There's another thing that I had in mind. I have also been reading on how PSO is implemented in other libraries and examples and was wondering if it would be a good idea to have a particle class.

09:40 < suryo> Instead of having the velocities and positions as variables within the main class itself, I was thinking that perhaps we could have a class for particles that could have the velocity and particles as member variables.

09:41 < suryo> Implementing this would probably take a tad bit longer than just moving the old code from the earlier versions of mlpack, but I think that by doing this, we could probably use the particles as templates, too.

09:42 < suryo> And then, based on the methods that we want to implement, in the future, while trying to develop a simple constrained version of PSO, we could try to feed the constraints into the individual particles

09:43 < suryo> and then make the particles move only towards feasible solutions.

09:44 < suryo> And of course, all that's required as an input is the number of particles, and we could have an array of particle objects within the PSO class, (the 'swarm' class)

09:44 < suryo> It's good from an object-oriented point of view, too.

09:45 < suryo> Let me know what you think about this. Thanks.

13:23 < zoq> suryo: Sounds reasonable to me, as long as we use armadillo matrices for the internal parameter, so that I can still use the existing interface to e.g. set the best particle e.g. iterate = bestParticle.coordinates.

21:29 < suryo> zoq: noted.