We are all familiar with the ability of neural networks to recognize handwriting. The basics of this technology have been around for many years, but it's only relatively recently that leaps in computing power and parallel processing have made it a very practical solution. However, this practical solution will, at its core, be represented by a digital computer that changes bits many times, just like it would when running any other program. But in the case of a neural network developed by researchers at the Universities of Wisconsin, MIT, and Columbia, things are different. They .
This glass contains precisely positioned inclusions such as air bubbles, graphene impurities and other materials. When light hits the glass, complex wave patterns occur, causing the light to become more intense in one of the ten areas. Each of these areas corresponds to a number. For example, below are two examples showing how light is propagated when the number "two" is recognized.
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 overcome the limitations caused by the glass production process. They started with a very limited design of the device to get a working prototype. Next, they plan to continue exploring 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 build a XNUMXD neural network in glass.
Source: habr.com