One step in artificial intelligence is to simulate how the human brain works on a computer.

Summary: There is still a long way to go before AI becomes more specialized than ever before as a human being can handle many tasks.

In recent years, AI has ushered in a huge leap forward. We have seen this technology applied to self-driving cars, collaborative robots, and multi-purpose deep learning systems (which can play all kinds of board games on their own, extrapolate routes using subway maps, or infer relationships using genealogical maps). However, there is still a long way to go before the AI ​​changes from a relatively specialized function to one that can handle many tasks as easily as humans.

A step toward the development of this strong artificial intelligence is to simulate how the human brain works on a computer so that researchers can better understand the underlying mechanisms behind intelligence. The problem is that the human brain is extremely complex. Even with the power of today's large supercomputers, it is impossible to simulate all the interactions between 100 billion neurons and trillions of synapses.

But now this goal is even closer, thanks to a group of international researchers who have now developed an algorithm that not only speeds up the simulation of human brains on existing supercomputers, but also to the next billion billion times. A "big brain" simulation has taken a giant step toward computing supercomputers (computers capable of performing billions of operations per second).

Total brain simulation required calculations

The study, published in the Frontiers in Neuroinformatics journal (https://?utm_source=G-BLO&utm_medium=WEXT&utm_campaign=ECO_FNINF_20180302_exascale-brain), outlines how researchers built a nerve on a supercomputer. New ways to network. To understand how daunting this task is, to illustrate just one example: Existing supercomputers (such as the Petascale Computing K Computer of the Kobe Institute of Advanced Computing Technology in Japan) can only replicate activities in 10% of the brain.

That is because it is constrained by the way in which the simulation model is built, which affects how the supercomputer's nodes communicate with each other. Supercomputers may have more than 100,000 such nodes, each with its own processor to perform operations. In larger simulations, these virtual neurons are distributed among computing nodes in order to efficiently balance the workload. However, one of the challenges of these larger simulations is the high connectivity of neural networks, which requires a lot of computing power. To copy.

Susanne Kunkel of the Royal Institute of Technology in Stockholm, Sweden, was one of the authors of the paper. She said: “Before neural network simulation can be done, we need to construct neurons and their connections in a virtual way, which means that they Need to create an instance in the memory of the node.In the simulation process, the neuron does not know at which node the target neuron is, so the short electrical pulse of the neuron needs to be sent to all the nodes, and then each node checks all these electrical pulses. Which are important for the virtual neurons that exist on this node."

In simpler and more visual terms, this is more than sending a whole stack of grasshoppers to each node, so each node needs to find the needles that are important to it from this stack of grasshoppers. Needless to say, this process consumes a lot of memory, especially when the size of the virtual neuron network increases. To use existing technology to scale and simulate the entire human brain requires 100 times more processing memory than today's supercomputers. However, the new algorithm changes the rules of the game because it can optimize the process. For this reason, it allows nodes to exchange which nodes send and which receive information, so that each node only needs to send and receive the information it needs. There is no need to look through the entire haystack.

The author of the paper, Jakob Jordan of Yulich Research Center, said: "With this new technology, we can make better use of the parallel mechanisms of modern microprocessors than ever before, which is in the billions of billions of times. Computing computers will become extremely important."

The research team found that with this improved algorithm, a virtual network of 520 million neurons connected by 5.8 trillion synapses is running on JUCHEEN, the supercomputer of the Lich Research Center, capable of operating at 5.2 minutes. Simulation of a biological time of 1 second, while the previous use of the traditional method requires 28.5 minutes of operation time.

It is predicted that machines capable of performing tens of billions of operations in the future will be 10 to 100 times more powerful than current supercomputers. With the help of the team's developed algorithm (to be provided as an open source tool), this will mean that the whole mechanism of intelligence can be explored.

Chlorinated Polyethylene Tape

Chlorinated Polyethylene Tape is also called CPE tape.

CPE plastic bag, it is made of high density polyethylene (HDPE) by chlorination substitution reaction of polymer material, not high temperature resistance, according to the structure and use of different, chlorinated polyethylene can be divided into resin chlorinated polyethylene (CPE) and elastic type of chlorinated polyethylene (CM) two categories. Thermoplastic resin can be used alone, but also with polyvinyl chloride (PVC), polyethylene (PE), polypropylene (PP), polystyrene (PS), ABS and other resins and even polyurethane (PU) blends.

It is translucent milky white, is made of high density polyethylene by chlorination substitution reaction of polymer material, feel more soft, not easy to deformation, aging resistance, not easy to wear.

Waterproof Electrical Tape HB1505 is a kind of Chlorinated Polyethylene Tape whitch has double layer,made from elasticity insulation materials and waterproof mastic.Overload temperature can reach 100℃ max.

Together with other Electrical Tapes,Special Adhesiveand and Insulating Tape,become our best-selling products.

Chlorinated Polyethylene Tape,Water Proof Tape,Waterproof Electrical Tape,Waterproof Sealing Tape

CAS Applied Chemistry Materials Co.,Ltd. , https://www.casac1997.com