European Human Brain Project leader Karlheinz Meier offers an overview of neuromorphic computing technology. He says a transition to such an architecture will make mimicking brain function not only more effective and efficient, but also will speed computational learning and processing beyond the pace of biological systems.
Among the milestones Meier cites is that available hardware systems have reached a high degree of maturity. Among the more prominent neuromorphic computing systems in use today is IBM's TrueNorth, which uses the TrueNorth chip deployed in complementary metal-oxide semiconductors. Since memory, computation, and communication are managed in each of its 4,096 neurosynaptic cores, TrueNorth bypasses von Neumann architecture bottlenecks and is highly energy efficient. This spring, IBM announced a joint project with Lawrence Livermore National Laboratory in which it will provide a scalable TrueNorth platform expected to process the equivalent of 16 million neurons and 4 billion synapses while only using 2.5 watts of power.
Leading neuromorphic machines face the formidable challenge of training neural networks, with Meier noting most currently used networks are deterministic. "The machines we are building at the moment are research devices but they have one important feature, they are really extremely configurable," Meier says. "In particular, you can also read out the activity of network because you want to understand what is going on."
From HPC Wire
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found