A developed circuit board can carry out deep learning computing!

“Deep learning” computer systems, based on artificial neural networks that copy the way the brain learns from an accumulation of examples, have become a hot topic in computer science.But the deep learning computations requires a highly complex and very powerful computers.

MIT researchers has developed a new approach to complex computing, using light instead of electricity, which they say could advance the speed and efficiency of certain deep learning computations. Their results appear today in the journal Nature Photonics in a paper by MIT postdoc Yichen Shen, graduate student Nicholas Harris, professors Marin Soljačić and Dirk Englund 

Soljačić says “people dramatically over-promised, and it backfired.” While many proposed uses of such photonic computers turned out not to be practical, a light-based neural-network system developed by this team “may be applicable for deep-learning for some applications,” he says.

“This chip, once you tune it, can carry out matrix multiplication with, in principle, zero energy, almost instantly,” Soljačić says. “We’ve demonstrated the crucial building blocks but not yet the full system.”

Shen says “The natural advantage of using light to do matrix multiplication plays a big part in the speed up and power savings, because dense matrix multiplications are the most power hungry and time consuming part in AI algorithms” he says

Traditional computer architectures are not very efficient when it comes to the kinds of calculations needed for certain important neural-network tasks. Such tasks typically involve repeated multiplications of matrices, AI which can be very computationally intensive in conventional CPU or GPU chips.

By way of analogy, Soljačić points out that even an ordinary eyeglass lens carries out a complex calculation on the light waves refracts. The way light beams carry out computations in the new photonic chips is far more general but has a similar underlying principle. The new approach uses multiple light beams directed in such a way that their waves interact with each other, producing interference patterns that execute the result of the intended operation (nanophotonic processor.)

To demonstrate the concept, the team set the programmable nanophotonic processor to implement a neural network that recognizes four basic vowel sounds. Even with this rudimentary system, they were able to achieve a 77 percent accuracy level, compared to about 90 percent for conventional systems. There are “no substantial obstacles” to scaling up the system for greater accuracy, Soljačić says.

Englund adds that the programmable nanophotonic processor could have other applications as well, including signal processing for data transmission. “High-speed analog signal processing is something this could manage” faster than other approaches that first convert the signal to digital form, since light is an inherently analog medium. “This approach could do processing directly in the analog domain,” he says.