UC Santa Barbara Researchers Create Functioning Artificial Neural Circuit
Researchers at the University of California,
Santa Barbara have created a circuit with about 100 artificial synapses and
demonstrated its ability to classify images.
As part of the demonstration, the artificial neural network successfully
classified images of the letters "z", "v" and "n", each of which was stylized
differently or saturated with visual noise.
The achievement is an early step toward the eventual goal of developing
artificial neural circuitry on the scale of a human brain, which contains one
quadrillion synaptic connections. According to information from UC Santa
Barbara, "the human brain remains a model of computational power and efficiency
for engineers" because it "can accomplish certain functions in a fraction of a
second what computers would require far more time and energy to perform."
The artificial neural network uses electronic components called memristors — a combination of memory and resistors — that change electrical
resistance depending on the direction of the charge. While conventional
transistors are based on the movement of electrons through semiconductors,
memristors are based on ionic movement, which is similar to the generation of
electrical signals in the human brain. According to information from the
university, memristors are much more efficient than transistors at creating
human brain-type functionality, but many more memristors will be required to
build more complex neural networks.
The next step for the researchers is to continue improving the performance
of the memristors and eventually integrate a memristor neural network with
conventional semiconductor technology in order to complete more complex tasks.
Practical applications for this technology could include improved medical
imaging, navigation systems and database searches based on images rather than
text.
The researchers have published their findings in the journal Nature.
About the Author
Leila Meyer is a technology writer based in British Columbia. She can be reached at [email protected].