The new algorithm will simulate the neural connections of the whole human brain on future exaflop supercomputers

In 2013, to simulate 1 second of operation of 1% of the human brain, it took 40 minutes on a cluster of 82,944 10-petaflops K computer processors . Scientists tried to replicate the work of 1.73 billion nerve cells and 10.4 trillion synapses connecting them, each of which took 24 bytes.

The power of new generation supercomputers will be calculated in exaflops, but with existing software solutions it will be enough to simulate only 10% of brain activity. An international team of scientists has created an algorithm capable of changing this and representing up to 100% activity. For the first time, researchers will have enough power of existing computers to simulate a neural network on the scale of the whole human brain.



The new simulation algorithm is designed to help simulate 100 billion interconnected neurons on exaflops supercomputers, that is, to “digitize” neurons on a whole brain scale. It is based on the NEST neurosimulation tool that the Human brain project works with . Researchers from several countries participated in the creation of the algorithm: the Julich Research Center (Germany), the Norwegian University of Life Sciences and Technology NMBU, Rhine-Westphalian Technical University Aachen (Germany), Institute of Physical and Chemical Research RIKEN (Japan), Royal Institute of Technology (Sweden).

“Before it is possible to simulate a neural network, neurons and their connections must be created in a virtual environment,” says study author Susanne Kunkel. During the simulation of 100 thousand neurons represented by the same number of network nodes, the excitation wave from the neuron should be sent to all 100 thousand nodes. Each of the nodes is equipped with processors for performing calculations. When receiving a signal, they are used, inter alia, to check its relevance - does this impulse apply to it.

Each signal corresponds to one bit of information per processor for each neuron of the network. For a network of a billion neurons, a large amount of node memory will be spent on relevance checking. As the network grows, this number grows, so to increase the share of activity in the simulation from 1% to 100%, it will require an increase in computer memory by 100 times. Since 2014, neural network simulations using NEST have been performed on K computer and JUGENE with a capacity of 10 petaflops and 222 teraflops, respectively. Future supercomputers will become more powerful, the number of processors for computing will increase, but the amount of memory per processor will remain the same.

The new NEST algorithm comes in handy here. At the beginning of the simulation, it will allow the nodes to determine what data on neural activity should be sent and where. Once this information is clear, it will be possible to send data in the address order. This will eliminate the need to process a data bit for each neuron in the network.



Software today is able to represent about 1% of the activity of neurons in the cerebral cortex on supercomputers, whose power is calculated in petaflops. The next generation of large-scale supercomputers will increase this figure to 10%. The new algorithm will make it possible to simulate up to 100% of brain activity at the same capacities using the same amount of memory.

Researchers are confident that after optimizing memory usage, the main task will be to increase the speed of simulation. For example, today, simulating 0.52 billion neurons, united by 5.8 trillion synapses, on a JUQUEEN supercomputer takes 28.5 minutes per 1 second of biological time. A new algorithm for calculating scientists will reduce this time to 5.2 minutes. “The combination of ex-scale iron and new software will allow us to explore the fundamental functions of the brain, such as plasticity and training, which take place over minutes of biological time,” says Markus Diesman. In 2013, the scientist claimed that the computing power needed to simulate the brain will become available after 2020 - in two years we will be able to find out if he is right.

In one of the next software releases from the Neural Simulation Technology Initiative, the code will be freely available to the entire community. The algorithm will speed up simulations on existing petaflops supercomputers, developers say.

Kenji Doya, an engineer and neuroscientist at the Okinawa Institute of Science and Technology, may be one of the first users of the new algorithm: “We worked with NEST to simulate the complex dynamics of the basal ganglia of a healthy person and Parkinson’s disease on K computer. We are pleased to hear about the new version of NEST, which will allow us to run a simulation of the whole brain on next-generation computers, which will make it possible to clarify the mechanisms of motor and mental functions. ”