Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
?
EE Times-Asia > Embedded
?
?
Embedded??

Using memristors to build an electronic brain

Posted: 05 May 2016 ?? ?Print Version ?Bookmark and Share

Keywords:e-brain? biological synapses? memristor? nanoelectronics?

Currently, Matveyev's team at MIPT is attempting to use its memristor-based E-Brain to mimic the common functions of real brains, such as memorising (called long-term-potentiation, or LTP, by neuroscientists) and forgetting (called long-term depression, or LTD, by neuroscientists), which is accomplished by changing memristor's connection strengths (also called weights since they are analogue values). LTP and LTD used together allow the brain to exhibits plasticitythat is, the ability to learn-new-tricks and to forget those no longer needed to maker room for new knowledge.

"The so-called "long-term potentiation" (LTP) and "long-term depression" (LTD) are the primary properties of biological synapses, which define their synaptic plasticity, that is the ability of synapses to change their weight (connection strength)," Matveyev told EE Times. "This property is believed to be the major cellular mechanism underlying learning and memory. LTP and LTD can be emulated in memristors as a gradual change of resistance"synaptic weight"in response to repeated pulsed biasing."

Even IBM emulates the way neurons stimulate synapses during learning using pulses (called spikes by neuroscientists) which work by counting the spikes received down its many inputs (called post-synaptic dendrites by neuroscientists) until they exceed a variable threshold number, then firing an analogue voltage spike down its single output (called an post-synaptic axon by neuroscientists) which is connected to hundreds and sometimes even though sands of other neurons throughout the brains (axons can be very long).

synaptic connections

Figure 3: The changes in potential of the real synaptic connections in real biological brains (left) compare favourably, but not yet exactly, to the changes in conductivity of MIPTs memristorsas a function of the temporal separation between "spikes" (right). (Source: MIPT)

"In our work, the voltage pulses from two different generators both have the shape of real neuron spikes (see photo)," Matveyev told us. "And are applied to the opposite electrodes with varying relative time delay, thus emulating spikes on pre- and post-synaptic neurons. The resulting spiking timing dependent plasticity function, which is the resistance change in the memristor as a function of relative voltage pulse timing, is similar to that displayed by biological synapses."

Many neuroscientists put great value in this "spike timing" function, claiming that the brain depends on simultaneous spiking to match sight, sounds, feelings, flavours and smells with the same objects. Without building this simultaneity into an E-Brain model, it could only learn individual properties rather than put all the properties into perceptions of multi-faceted objects as humans do. This simultaneity also accounts for the brains incredibly fast operation, to solve complex problems, while consuming only 20 wattssince neurons fire only 10 or so time per second.

- R. Colin Johnson
EE Times


?First Page?Previous Page 1???2



Article Comments - Using memristors to build an electro...
Comments:??
*? You can enter [0] more charecters.
*Verify code:
?
?
Webinars

Seminars

Visit Asia Webinars to learn about the latest in technology and get practical design tips.

?
?
Back to Top