Dec 11, 2017 | Updated: 09:54 AM EDT

Deep Learning Computer Is Faster With Light, Instead Of Electricity

Jun 13, 2017 05:19 PM EDT

Close
Next gen laser tag using AI machine learning, 3D printed weapons and drones that can shoot at you

A new photonic technology has enabled a computer system to mimic the way human brains learn from accumulating experience. Researchers from the Massachusetts Institute of Technology has developed a new approach for the deep learning computation using light, instead of electricity.

A deep learning computer is a way computer system accumulate experiences and data and recognized the pattern in the accumulative data. Unfortunately, even the most powerful computer is limited with its transistor capacity to perform such function. In order to improve the deep learning computer system, researchers from the Massachusetts Institute of Technology discovered that light is a much better answer to perform such function, instead of electricity.

Their research has been published in the Nature Photonics Journal. Lead author of the paper is a post-doctoral researcher from Physics Department at the MIT, Yichen Shen and a Ph.D. candidate, Nicholas C. Harris. Shen performed the experiment while the photonic chip was developed and built by Harris, and nine others researchers also contributed in the study.

The new photonic computer system provides the answer for all possible practical applications for the deep learning computer system. Many possible applications for this photonic computer are data centers, security systems, self-driving cars and other applications that require a lot of computation which require a huge power or time. Using light to do matrix multiplication for this deep learning computer system saved a huge amount of power, thus increase the speed of processing.

“The natural advantage of using light to do matrix multiplication plays a big part in the speed up and power savings," Shen said about the deep learning computer system using light. "Because dense matrix multiplications are the most power hungry and time-consuming part in AI algorithms.”

In designing the new programmable nanophotonic processor, Harris used an array of waveguides and interconnected them so they can be modified as needed with a program that direct the waveguides in specific computation. When the program is executed, the processor will guide the light to series of coupled photonic waveguides.

©2017 ScienceTimes.com All rights reserved. Do not reproduce without permission. The window to the world of science times.
Real Time Analytics