Neural network in the glass. Does not require power, recognizes numbers

Original author: hackaday.com
  • Transfer
image

We are all familiar with the ability of neural networks, such as handwriting recognition. The foundations of this technology have been around for many years, but only relatively recently, a leap in the field of computer capacities and parallel data processing made it possible to make a very practical solution out of this technology. Nevertheless, this practical solution, basically, will be presented in the form of a digital computer repeatedly changing bits, in the same way as when executing any other program. But in the case of a neural network developed by researchers from the universities of Wisconsin, MIT, and Columbia, things are different. They created a glass panel that does not require its own power supply, but at the same time capable of recognizing handwritten numbers .

This glass contains precisely located inclusions, such as air bubbles, impurities of graphene and other materials. When light hits the glass, complex wave patterns arise, resulting in light becoming more intense in one of ten areas. Each of these areas corresponds to a number. For example, two examples are shown below, showing how light travels when recognizing the number two.

image

With a training set of 5000 images, the neural network is able to correctly recognize 79% of 1000 input images. The team believes that they could improve the result if they could get around the restrictions caused by the glass production process. They started with a very limited device design to get a working prototype. Further, they plan to continue studying various ways to improve the quality of recognition, while trying not to overly complicate the technology so that it can then be used in production. The team also has plans to create a three-dimensional neural network in glass.

Also popular now: