The Wall Street Journal (03/28/16) Robert McMillan
The Lawrence Livermore National Laboratory (LLNL) on Thursday will begin testing a $1-million computer packed with 16 IBM TrueNorth microprocessors designed to mimic the functions of the human brain. Bundled into each TrueNorth chip are 5.4 billion transistors comprising a network of 1 million simulated neurons connected by a massive web of synapses. "TrueNorth is useful for deep-learning applications and for a broader class of machine-learning applications as well," says LLNL researcher Brian Van Essen. TrueNorth emulates the brain's low power consumption, with the 16 chips using only 2.5 watts together versus a typical server chip's power requirements of up to 150 watts. Van Essen's team will test TrueNorth by uploading some supercomputing tasks to it. Van Essen expects the system to help the lab filter out potential glitches in simulations of phenomena such as subatomic particle interactions and identify patterns in cybersecurity and video surveillance. "It's great that they're [testing TrueNorth]," says University of Washington professor Luis Ceze. "It's very efficient, but they have to show that the accuracy of the models that they implement [is] good enough."
Aucun commentaire:
Enregistrer un commentaire