Sign In

Communications of the ACM

ACM TechNews

New System Allows Optical 'Deep Learning'


Artist's depiction of programmable nanophotonic processors integrated on a printed circuit board.

A team of researchers has developed a new approach to deep learning, using light instead of electricity.

Credit: RedCube Inc., Y. Shen, N. Harris, M. Solja?i?, D. Englund, et al.

Researchers at the Massachusetts Institute of Technology (MIT) have developed a new approach to deep-learning computer systems that uses light instead of electricity, which say they could vastly improve the speed and efficiency of certain computations.

Although other researchers have proposed uses of photonic computers that ended up being impractical, the researchers say their light-based neural network system may be applicable for deep learning for some applications.

"We've demonstrated the crucial building blocks but not yet the full system," says MIT professor Marin Soljacic.

He notes the optical chips using this architecture could, in principle, execute calculations conducted in typical artificial intelligence algorithms much faster and using far less energy for each operation than conventional electronic chips.

The programmable nanophotonic processor uses an array of waveguides interconnected in a way that can be modified as needed, coding that set of beams for a specific computation.

From MIT News
View Full Article

 

Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


 

No entries found