IBM have reproduced the activity of neurons using devices with phase transition

Researchers at the IBM Research in Zurich have created artificial neurons on the basis of devices with phase change, which behave like neurons in a living organism. According to scientists, this is a significant step in the development of energy-efficient neuromorphic integrated solutions built on the principles inherent in the human brain. It is assumed that they will be in demand in cognitive applications, that is associated with obtaining, processing and use of knowledge.

Modern computers are inferior to «biological computers» on density layout and energy efficiency, so they are poorly suited to simulate the stochastic, that is, possessing a certain degree of randomness, in the brain, including a large number of neurons.

Neurons on the basis of the devices with phase transition implement the functions of storing and processing data, in particular, can determine the degree of correlation data and self-study with high speed with very little energy cost.

Even a single neuron can perform a specific function — it changes its state after some number of pulses. Combining neurons, it is possible to implement mechanisms for the analysis of sensors in the Internet of things or identifying patterns in financial data.

Source: IBM



(Visited 16 times, 1 visits today)
No tags for this post.