search

psycharo-zz / factor-graph

matlab/c++ factor graph framework
32 stars 8 forks source link

speed up #55

Open psycharo-zz opened 11 years ago

psycharo-zz commented 11 years ago

training a model with 32 mixtures currently takes around 1.5 seconds. since networks are entirely independent, it should be possible to multithread (c++11 ThreadPool implementation - https://github.com/en4bz/ThreadPool) Also, when fitted, a large number of mixtures has very low weight (those components with mean=0,prec=1,weight<~1e-5), however they decrease performance. they could be easily removed after pre-training, or even online somewhere in vmp::MoGArray.

psycharo-zz commented 11 years ago

http://arma.sourceforge.net/speed.html - using this library instead of std::vector might also improve performance