The scientists – at the University of California Santa Barbara – said that it tested a circuit of 100 artificial synapses to match a simple task of image classification.
Dmitri Strukov, a professor of electrical and computer engineering, described it as a small but important step towards scaling up to the functions of your brain.
Strukov said that the human brain is a model of computational power and efficiency and can do things in a fraction of a second, while computers need far more time and energy.
The test was able to classify three letters of the alphabet – z, v and n – by their images, with the letters being stylised in different ways or swamped with noise.
The scientists use memristors – components where resistance changes according to the direction of the flow of the electrical charge.
But it’s early days for the tech, the scientists warned although there are potential applications including medical imaging, and navigation systems.
The next step in the quest for the neural network holy grail is to integrate memristor neural networks with conventional semiconductor technology.