Industrial & Medical Technology

Better AI Through Optics

19 July 2018
Illustration of a neural network trained by an optical circuit. Laser inputs (green) encode information carried through the circuit (blue) by optical waveguides (black). During training, beam splitters (curved sections) can be tuned by adjusting the settings of optical phase shifters (red and blue glowing objects). Source: Tyler W. Hughes/Stanford University.

Artificial neural networks are a type of artificial intelligence that uses connected units to process information, similar to the way our brains work. By training the algorithms to categorize inputs, they can perform complex tasks. This type of processing is typically performed with a conventional computer, but significant efforts are under way to design optimized hardware that can increase efficiency. That’s where optics-based devices, which can perform parallel computations while using less energy than electronics, come in.

Researchers at Stanford University have just reported a breakthrough using an optical circuit to train an artificial neural network, which could lead to less expensive, faster and more energy-efficient ways to perform tasks like voice or image recognition.

"This would enhance the capability of artificial neural networks to perform tasks required for self-driving cars or to formulate an appropriate response to a spoken question, for example,” noted research team leader Shanhui Fan. “It could also improve our lives in ways we can't imagine now."

By designing an optical chip that replicates the way that conventional computers train neural networks, the researchers overcame a significant challenge to implementing an all-optical neural network. The new training protocol operates on optical circuits with tunable beam splitters. Laser beams encoding information to be processed are fired into the circuits and carried through the beam splitters by optical waveguides. These can then be adjusted via optical phase shifters, which function like knobs.

The technique was successfully tested with optical simulations in which an algorithm was taught to perform complicated functions, such as picking out complex features within a set of points.

"Our work demonstrates that you can use the laws of physics to implement computer science algorithms," said Fan.

The general approach designed by the team could be used with various neural network architectures, and for other applications such as reconfigurable optics.

The research was reported in Optica, the journal of The Optical Society.

Powered by CR4, the Engineering Community

Discussion – 0 comments

By posting a comment you confirm that you have read and accept our Posting Rules and Terms of Use.
Engineering Newsletter Signup
Get the GlobalSpec
Stay up to date on:
Features the top stories, latest news, charts, insights and more on the end-to-end electronics value chain.
Weekly Newsletter
Get news, research, and analysis
on the Electronics industry in your
inbox every week - for FREE
Sign up for our FREE eNewsletter