Please activate JavaScript!
Please install Adobe Flash Player, click here for download
ePaper created 2013-08-02, 08:40:53 | version 1.25.20
112|2013 Research in Jülich The exchange between neuroscientists and IT spe- cialists is based on the principle of give and take. For example, medical scientists use the super- computers in order to improve their understanding of the brain. However, the experts at the Jülich Supercomputing Centre (JSC) also benefit from the new findings on the brain – for example when it comes to energy efficiency: while large computing centres consume enormous amounts of electricity, the brain requires only about the same amount of power as a weak incandescent bulb. In addition, the human brain tolerates a lot more errors than a supercomputer. Although neurons die in the course of a lifetime, the brain continues to function in healthy humans. This is quite different in super- computers: if one processor malfunctions, the re- maining 100,000 processors may also be switched off. The brain therefore serves as a model for Jülich’s IT specialists. “New findings on the function of the brain can inspire us to find new options for data pro- cessing and incorporate them in the development of new generations of computers,” says Prof. Thomas Lippert, director at JSC. One of the creative thinking spaces for these new computer architectures is the Simulation Laboratory Neuroscience, which is head- ed by physicist and neuroscientist Prof. Abigail Mor- rison. “Our computer experts are working closely with their colleagues in neuroscience. They support them in ‘translating’ medical questions into a lan- guage the computer is able to calculate, and in this process, they are learning more every day about the brain, our ingenious information network.” :: COVER STORY | Human Brain Modelling Using the Brain as a Model Live 3D Simulation In order to be able to map the 90 billion neurons in the brain in the future, supercomputers of the next generation will have to provide a computing power that is 1,000 times higher than that of to- day’s supercomputers, and will also need a high storage capacity and storage bandwidth. At the same time, new simulation techniques will be re- quired in order to be able to effectively exploit this computing power – a task taken care of by the research group ‘Computational Neuro- physics’ headed by Prof. Markus Diesmann. There are also completely new challenges in terms of what is simulated, explains Dr. Boris Orth from the Jülich Supercomputing Centre, who works closely with teams of neuroscientists every day. “We are currently running simulations on our computers that represent the structure and workings of individual areas,” says Boris Orth. Neuronal fibre tracts, molecular mechanisms, or spiking activities: the supercomputers simulate clearly defined tasks. The future, however, belongs to ‘interactive supercomputing’, which incorporates all the knowledge available, irrespective of whether it’s molecules interacting, individual brain areas communicating, or whether it’s the entire brain that’s active. It sounds like something out of Star Trek: in a few years’ time, researchers hope to be able to watch arithmetic operations in the virtual brain as they happen, in three dimensions on huge displays in a type of ‘mission control cen- tre’. “On the one hand, the huge challenge con- sists in integrating data on various levels into one single simulation,” says Boris Orth. On the other hand, simulation and visualization must take place in parallel. The advantages are obvi- ous: while the simulation is running, the re- searchers can zoom in on certain areas, even the molecular level, as with a telescope, in order to change simulation settings there if necessary. In this way, it would be possible to observe and test the effects of drugs and other treatments such as neurostimulation directly in the brain – which would be a milestone for drug development and brain researchers comparable to the first step on the moon. :: Thomas Lippert, physicist Boris Orth, physicist Ilse Trautwein JSC