Electronic Design

Neuromorphic Engineers Craft Brain-Like Circuitry To Perform Perceptual Tasks

Scientists at Bell Labs and the Massachusetts Institute of Technology have developed an electronic circuit that mimics the biological circuitry of the cerebral cortex. In the future, devices that behave like neural systems could be able to perform complex perceptual tasks, such as recognizing objects by sight. Researchers call this field neuromorphic engineering.

The circuit is composed of artificial neurons that communicate with each other via artificial synapses. These elements are made from transistors fabricated on a silicon IC. Like the neurons in the cortex, these artificial neurons affect other nearby neurons. An inhibitory neuron receives input from the 16 excitatory neurons that also reside on the circuit. By returning inhibition back to the excitatory neurons, the inhibition neuron stabilizes the circuit by keeping excitatory feedback in check.

When two artificial neurons receive simultaneous electrical currents, the circuit responds to one stimulus while repressing the other. No single element decides which stimulus gets suppressed. Instead, the decision results from an emergent, collective property of all of the neurons. This is very similar to what happens in the human brain.

A human neuron is typically connected to 10,000 other neurons. The brain contains billions of these neurons in a vast, complex network, creating a tangled web of feedback loops. Yet the brain doesn't use feedback in the strictly analog or strictly digital way that conventional electronics use them. Instead, the brain combines these functions.

Perception, researchers say, depends on analog and digital properties. When a person sees a car, for example, the brain receives a steady stream of analog information. What color is it? How big is it? How fast is it going? At the same time, the brain makes a simple digital decision—is it a car or not?

Research suggests that these processes aren't exclusive, either. Scientists say the brain's neural circuitry appears to be a hybrid where analog and digital processes coexist. Managing this dual nature could be the key to someday creating devices with human-like perception.

Much of the work was done at Bell Labs in Murray Hill, N.J. The circuit was designed in collaboration with the Institute of Neuroinformatics in Switzerland. The Swiss National Science Foundation, Lucent Technologies, and MIT supported the effort. For more information, contact Steve Eisenberg at Bell Labs at (908) 582-7474.

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish