“Largest neuronal network simulation achieved using K computer”

[Computational Neuroscience]

In our brains, information in the form of electrochemical signals is processed and transported from one neuron to the next at speeds up to 250 miles per hour via connectors called synapses. We’re information processors—this is why computer metaphors are sometimes apt for describing our brains. But how true is this metaphor? “K computer” at the Okinawa Institute of Technology Graduate University (currently the 4th fastest computer on Earth) is getting closer to answering that question. With the processing power of 250,000 high speed PCs, K computer has performed the largest simulation of a neural network ever. The researchers simulated the activity of 1.73 billion neurons connected by 10.4 trillion synapses. Using 82,944 processors, K computer took 40 minutes to simulate 1 second of random brain activity.  While this is nowhere near as fast as a human brain, nor is the network even close to as large (human brains are estimated to have about 100 billion neurons), this is an important step in understanding neural networks. It also opens up vast research spaces to test the limits and boundaries between neural and computer networks. Is this a science fiction writer’s dream, slowly—very slowly—becoming a reality?

Tags: , , , , , , ,