US, WASHINGTON (ORDO NEWS) — Intel Corporation announced the launch of its largest neuromorphic computer. The system contains one hundred million neurons – more than the brain of a house mouse.
Recall that such machines mimic the structure of the brain. The mentioned chip contains 130 thousand artificial neurons and 130 million interneuronal connections (synapses). It solves some specific tasks a hundred times faster than traditional computers of comparable computing power.
Now, engineers have combined 768 Loihi instances into a single neuromorphic Pohoiki Springs computer. It is easy to calculate that the new system has about one hundred million neurons.
This is more than in the brain of a house mouse (71 million nerve cells) and is comparable to the brain of a gray rat (200 million). Meanwhile, we recall that rats are famous for their quick wits.
At the same time, the device turned out to be quite compact: Pohoiki Springs takes up no more space than five standard servers. Power consumption is also quite modest: only 500 watts.
The system will be available for cloud computing to Intel Neuromorphic Research Community members. For this, the corporation provides appropriate software.
The development is intended in order to assess the potential of neuromorphic computers in solving a wide range of tasks requiring high computing power.
In particular, many problems in various fields, from logistics to chemistry, can be reduced to finding the optimal path on the graph. Recall that a graph is a set of points connected by lines.
For example, settlements connected by roads, on the map represent a graph. The scheme of family ties between people, interactions between chemicals and much more can also be represented in the form of a graph.
Classical for graph theory is the task of the Chinese postman: how to get around all the streets of the city and return home, having done as little as possible? It is clear that similar issues arise, for example, in front of transport companies.
Meanwhile, a exhaustive search of all route options is usually unbearable even for the most powerful computers.
For many years, mathematicians have been working on algorithms that optimize the search for an answer. Now the researchers intend to test how effective the approaches developed specifically for the neuromorphic computer will be.
The search for the optimal path on the graph is just a special case of the problem in which you need to achieve the best result, taking into account some restrictions.
Other examples of such tasks are planning a flight schedule or, say, a company’s budget. Here, too, Pohoiki Springs can prove itself.
Specialists expect to apply the new system to such traditional for neural networks areas as pattern recognition and image search for certain structures.
Contact us: [email protected]
The article is written and prepared by our foreign editors from different countries around the world – material edited and published by Ordo News staff in our US newsroom press.