Computer hardware designed for replicating human brain

Computer hardware designed for  replicating the human brain

Researchers at the University of Sussex have made the fastest and most energy-efficient simulation of the brain part of the brain using off-the-shelf computer hardware.

Dr of the Engineering and Informatics of the University of Sussex James Knight and Professor Thomas Nowotny beat the top 50 supercomputers by running brain simulations using their own GeNN software and graphics processing unit (GPU).

By developing faster and more efficient simulators, academics are expected to increase the level of understanding in brain function and, in particular, recognize that damage to specific structures in neurons can cause brain reduction.

 Faster, more advanced simulators can help improve the understanding of neurological disorders by pointing to those areas of the brain that cause epileptic seizures.

read more: Artificial intelligence may help reduce gadolinium dose in MRI

Advanced simulators can accelerate progress within the development of AI - JNN software is being used to build autonomous robots with flying drones already in Sussex University, which can be controlled through fake insects.

Prof Nowotny, professor of information science at the University of Sussex, said: "In the last three decades, computers have become much more powerful, due to our ability to create computer chips with small and small components at large.

 which in return This process allows a wall to be killed and it is fundamentally different from other architectures without a fast computer. 

The GPU is such an architecture and our work shows that in the near term, they are a competitive design for high-performance computing. As far as the CPU has moved us so far, There is the ability to move beyond that.

The research involved using the team's own GeNN software to implement and test two established computational neuroscience models; Of the cortical microcircuit.

 Eight population of neurons and spike-time dependent plasticity is a balanced random network - a process that is shown to be fundamental for biological learning.

A single GPU was able to accelerate processing speed up to 10%, currently using supercomputer or SpiNNaker neuromorphic system, £ 1bn developed as a part of the European Human Brain Project (Hbp), a custom-built machine.

The team of Sussex University was able to save 10 times the power of SpiNNaker or Super Computer simulation.

Going forward, academics believe that the flexibility and power of the GPU mean that they can play an important role in creating simulators capable of running models that begin to approach the human brain's complexity.

Research fellow at Computer Science at Sussex University Dr. Knight said: "Although we are going a long way from keeping the understanding of the entire human brain necessary, we are contacting at the point where the latest exclusive supercomputers are the raw computing power that emulates them Many of these systems are dependent on GPUs.

So we are happy with these latest results which show that the brain simulates How well suited GPU. Next year we will work 50 times as large model of a monkey visual systems using are expected expansion. Multiple interconnected GPU. "

Chris Emerson, chief of high-education and research sales in the UK and Ireland in NVIDIA, said: "We are very impressed with the use of NVIDIA AI Computational Platform for spearheading and brain simulation at Sussex University, and we are happy that we are working on computational neuroscience Able to support research on leading edge as well as AI. "