As a dabbler in machine learning, I understand that this takes an ginormous amount of resources and hard work to accomplish!
It is beyond amazing to me that a random virtual model matched 98% of the locations of synapses.
That is incredible accuracy for all the different types of classifications they showed us. These computations for the Blue Brain project were done on IBM’s BlueGene super computer. Currently, I am studying Thread Building Blocks (TBB) that was developed by IBM to do parallel processing with ease. I’d be willing to bet that they use TBB in their modeling architecture for Blue Brain.