Electrically tuneable network learns fast – Physics World“With this model, we can search for a desired functionality,” Van der Wiel says. “In standard deep learning, we have to find the parameters of the model itself,” he says. “With this new method we can optimize nanoelectronics devices with many terminals and even optimize systems in which many complex nanoelectronics circuits are coupled,” Van der Wiel explains. “Such systems are expected to increase in complexity in the coming years in, for example, novel information processing technologies like quantum computing and neuromorphic computing.” “While not as slow as Darwinian evolution, it is still quite time-consuming to have the network do what you would like it to do,” he tells Physics World. To generate a DNN model of their nanoelectronic device, Van der Wiel and colleagues began by measuring the device’s output signal for many distinct input voltage configurations. As a next step, the team plans to build more energy-efficient large-scale systems of interconnected dopant network devices for state-of-the-art AI performance.
Article from physicsworld.com.