Compared with IBM recently released for its cognitive computer digital hardware simulation components, Moscow Physics Institute of Technology (Moscow Institute of Physics and Technology; MIPT) researchers to use the analog components to build an electronic Brain (E - Brain), allowing it to work like a real biological Brain.

Russia's electronic brain key components will use today - this is found by chua, a professor at the university of California, Berkeley,; But there is also a pioneer, is the result of neural network and Stanford university professor Bernard Widrow. Anyway, today, just as its name implies, it can through the change of resistance memory how many current flow and its flow. MIPT and further to have resistance to 40 x40 nanometers. ?

MIPT using hafnium oxide (HfO2) synaptic gap (bright) complete analog electronic brain (source: MIPT)

In the human brain, the important functions are performed by the continued growth of the chemical "synaptic" - because the synaptic growth continuously, make it to reduce its resistance at the same time, also improved the conductivity. If it is not used, synapse will shrink until conductivity to zero. This function is usually start during sleep at night, in order to strengthen the important events in the daytime. During the day, people may be because of preparing for the examination and learning (cognitive memory), practice some sports (muscle memory) or memory words, phrases or list (the image memory), and makes some synaptic growth.

As early as 2011, HP (Hewlett Packard HP) for today is to build an electronic brain had high hopes, but later because of its attention to and Hynix (Hynix) to jointly develop commercial applications have lockout mega-ohm resistor chip; However, startup Knowm recently released today is based processor components.

Electronic "surge" signals transmitted through brain neurons to another (black line), and other various biological signals synchronous transmission (source: MIPT)

Today, supported by the Russian government MIPT picked up the baton, full build based on the analogy principle of electronic brain, allowing it to like a real brain, not only can behave more wisdom and more energy-efficient than traditional computer. MIPT general materials using CMOS fabs - hafnium oxide (HfO2), to develop the gen today, and bring the important innovation. Now, MIPT one by one for its consolidation, and ultimately create analog electronic brain, and expect to go beyond the IBM digital "cognitive computer".

"We first in embedded by a thin layer of metal oxide (especially the HfO2) today on the applied voltage, can drive from the lattice oxygen ions and further to one of the electrodes, left (charge) oxygen vacancy defects, and provide the electronic transmission path," MIPT nanoscale electronic functional materials and components laboratory, a senior fellow at Yury Matveyev explained, "so, today is the resistance value of defined by filamentous passage of oxygen vacancy concentration, and most importantly, it will offset the device today during change with time."

At present, in the MIPT Matveyev's team is trying to use the basis of today's electronic brain, simulate the actual human common functions, such as memory (i.e., a neuroscientist at the so-called long-term gain effect or LTP) and forgetting (neuroscientist so-called long-term inhibition effect or LTD), is by changing strength of today connection (because is a kind of ratio, thus also known as weight). LTP and LTD mutual collocation, let the brain plasticity, learning new skills and forget things is no longer needed in order to learn new knowledge and ability to give more space.

"So-called long-term gain effect (LTP) and long-term inhibition effect (LTD) is the main feature of biological synapses, define its synaptic plasticity - the synapse's ability to change its weight (connection strength)," Matveyev pointed out that "this feature is considered to be important mechanism of learning and memory cells. In order to respond to repetitive pulse bias, LTP and LTD can be simulated to have mega-ohm resistor of resistance -" synaptic weights "(synaptic weight) - a gradual change."

Real practical synaptic connections in biological brain (left) potential changes, compared with MIPT conductance coefficient change of today, as a temporary area lies between the function of the "surge" (right) (source: MIPT)

IBM while the pulse simulation neuron synapses during learning approach, the principle of which is calculated to receive surge and the input variable threshold until more than code, then start the analog voltage surge to the single output (i.e., the synapse rear tu), make it to connect to the many other neurons in the brain (some axons may long).

"In our study, from two different generator voltage pulse can see real neurons surge shape (pictured)," Matveyev explained, "and on the opposite electrodes with different relative time delay, so as to simulate the synapse neurons before and after the surge. The result was a surge in time depends on its plasticity functions, as have the resistance of a resistance change as the function of the relative voltage pulse timing, similar to the biological synapses show behavior."

Many neuroscientists attaches great importance to the value of "surge occurred time", said the brain depends on simultaneous surge, for the same thing with sight, sound, feel, taste and smell, etc. If unable to create this kind of synchronicity, electronic brain model then can only learn to separate features, not like people to integrate all the features into the various objects perception. This synchronization features also explains why the brain can so quick operation and solve complex problems, and the power consumption is only 20 w - because neurons discharge can be as low as 10 times per second.