Optical neuromorphic processor processes 2 billion images per second

(ORDO NEWS) — Researchers at the University of Pennsylvania have developed a powerful new optical chip that can process nearly 2 billion images per second.

We have never seen such speed before. Could this technology take computer computing to the next level?

The basis of the new chip is a neural network – a system that processes information like a brain. These networks are made up of nodes that are interconnected like neurons and even “learn” in a similar way to an organic brain by learning from datasets.

They can, for example, recognize objects in images or words in speech. Over time, they become much better at solving these problems.

Optical brain

But instead of electrical signals, the new chip receives and transmits information using light. As neurons, he uses optical wires stacked in several layers, each of which specializes in a certain type of classification.

The collection of these neurons can be used to solve complex problems – for example, recognizing letters and placing them in the desired sequence.

The scientists made a 9.3mm2 chip and tested it with a simple task: classifying a series of handwritten characters that looked like letters.

After being trained on the appropriate data sets, the chip was able to classify images with 93.8 percent accuracy for sets containing two types of characters, and 89.8 percent accuracy for four types.

Most impressively, it took the processor 0.57 nanoseconds to classify a single character. This means that it can process 1.75 billion images per second.

According to scientists, the new chip owes this speed to its ability to transmit information at the speed of light. Therefore, it has a great advantage over traditional computer processors.


Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.