K computer is unleashed for largest neuronal network simulation ever

by Nancy Owano weblog
The K computer
The K computer

(Medical Xpress)—Wow, today's top supercomputers have formidable speed and computational power to rival humans. On the other hand, wow, the human brain is more powerful than today's supercomputers, which with all their calculations, cannot surpass the complex wonder called the human brain. Scientists note that the brain, the basis of behavior and mental activity, is where over 50 billion neurons communicate with each other, in a network of complexity that some called "unimaginable." Each individual neuron's performance depends on molecular structures and mechanisms, and connections and spatial location in the brain.

A team of scientists from institutions in Japan and Germany have proven both "wows" are true. A report in this week's The Telegraph discusses the feat of a team from Japan and Germany who carried out a record simulation of the human brain, in the largest neuronal network simulation to date—and yet achieved a model of only one second of human brain activity, and taking the computer 40 minutes to do so. Consider: a single second's worth of activity took one of the world's largest supercomputers 40 minutes to calculate, and involving just one per cent of the neuronal network in the human brain. The joint project was carried out by the Japanese research group RIKEN, the Okinawa Institute of Science and Technology Graduate University and Forschungszentrum Jülich, a Germany based research center,

The team used RIKEN's K computer, which has its home in Kobe, Japan, known for exceptional precision and and also ranked as one of the world's most-powerful supercomputers, to simulate human brain activity. (The K computer was officially made available for use by researchers in September 2012.)

The team made use of the open-source Neural Simulation Technology (NEST) tool. The NEST project site describes it as a simulator ideal for networks of spiking neurons of any size. NEST was put to work to replicate a network of 1.73 billion nerve cells and 10.4 trillion synapses.

While only a second's worth of was achieved in 40 minutes, the project team nonetheless values their effort, as the goal was not to compete with or discover something entirely new about the brain but rather to test simulation technology and the capabilities of the K computer.

According to RIKEN's announcement, the researchers came away with knowledge that can guide the construction of new simulation software. Their work, said RIKEN, also "offers neuroscientists a glimpse of what can be achieved by using the next generation of computers—so-called exascale computing."

Markus Diesmann of Forschungszentrum Jülich's Institute of Neuroscience and Medicine, said, "If petascale computers like the K computer are capable of representing one percent of the network of a today, then we know that simulating the whole brain at the level of the individual nerve cell and its synapses will be possible with exascale computers—hopefully available within the next decade."

More information: www.telegraph.co.uk/technology… -brain-activity.html
www.riken.jp/en/pr/press/2013/20130802_1/

Related Stories

Blue Gene active storage boosts I/O performance at JSC

Nov 18, 2013

The ever growing complexity of various simulation tasks not only require a continuous increase of computing power but also the capability of managing large amounts of data. Realisation of an active storage ...

New theory of synapse formation in the brain

Oct 10, 2013

The human brain keeps changing throughout a person's lifetime. New connections are continually created while synapses that are no longer in use degenerate. To date, little is known about the mechanisms behind ...

Recommended for you

Fruit fly lights up brain wiring

2 hours ago

(Medical Xpress)—Fluorescent fruit flies have helped University of Queensland researchers take a critical step toward understanding the human brain's neuronal "wiring" and how it can go awry.

User comments

Adjust slider to filter visible comments by rank

Display comments: newest first

Huns
not rated yet Jan 14, 2014
Please proofread your stories so that we don't have to read the same sentence twice or wince when we see stray letters.
Horus
not rated yet Jan 14, 2014
I prefer the implied notion of `flinch' over wince. It's sudden and non-lasting, whereas when one winces in pain it's not a quit jarring reaction to the body or mind.
Zera
3 / 5 (1) Jan 14, 2014
However, if I am interpreting Kurzweil correctly (not that he is the last word on the subject either), it is not neccessary to simulate the entire brain, given as how many information channels are redundant, repitition, etc.

What I find more interesting is the power consumption, I imagine K Computer chews through a fair few kWh.

Interesting that it takes 40 mins to do 2 secs, it gives a benchmark from which to move foward, so as cpu's/gpu's chase higher and higher bandwidth/mem, etc.... now supercomputers or computer networks will chase human brain replication....
Kedas
2 / 5 (1) Jan 15, 2014
Current computers are too accurate to reproduce the brain (without a lot of power)
alfie_null
1 / 5 (2) Jan 15, 2014
I sense a certain amount of hyperbole in the announcement. What two seconds of brain operation were simulated? It's not as if you can "turn on" any given brain and it starts running. A brain is the product of a lifetime's accumulation of experience. And, as I understand it, we are a ways from knowing how all the components of a brain are connected. Or what influences those components as the brain operates. Hence, we aren't able to accurately simulate them.
210
5 / 5 (1) Jan 15, 2014
I sense a certain amount of hyperbole in the announcement. What two seconds of brain operation were simulated? It's not as if you can "turn on" any given brain and it starts running. A brain is the product of a lifetime's accumulation of experience. And, as I understand it, we are a ways from knowing how all the components of a brain are connected. Or what influences those components as the brain operates. Hence, we aren't able to accurately simulate them.

"...While only a second's worth of brain activity was achieved in 40 minutes...? A second...see? Two seconds would probably have been twice as long a runtime. On a machine of this size AND CLASS, even a millisecond of runtime is world-class expensive, try to appreciate that. No you cannot 'turn on' but you can record and then analyze any ONE second period of activity...download NEST and play with it yourself...then re-read this article.

word-
Sean_W
not rated yet Jan 15, 2014
One nice thing about comparing human brain capacity with technological metrics is that artificial intelligences won't need to meet human brain capacity since much of the brain is used for things like monitoring bowel function, obsessing over sex, learning how to avoid mosquitoes keeping track of which foods makes one feel bloated. A lot of our brain power is taken up by biological concerns.
Osiris1
not rated yet Jan 17, 2014
Might add to Sean's post, accurate though it is about how men use their brains, to post that women also waste brain function: obsessing about whether their 'other' is 'messing around', finding a parking place when shopping, gossiping about that 'brazen hussy' that just moved into their trailer park, etc, getting enraged over Mrs Rosenstern who has the best figure in town and the other women cannot lose the weight to compare with her........or maybe trying to attract the attention of that 'cute pool cleaner guy'.......

Also, it might be said that many of us are not all that intelligent, so some optimization of our own brains might be within the capabilities of future AI studies and its relevant hardware products that may result.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.