New article demonstrates the use of photonic structures for AI

As we enter the next chapter of the digital age, data traffic continues to grow exponentially. To further improve artificial intelligence and machine learning, computers will need the ability to process large amounts of data as quickly and efficiently as possible.

Conventional methods of computing are not up to the task, but when looking for a solution, researchers saw the light – literally.

Light-based processors, called photonic processors, allow computers to complete complex calculations at incredible speeds. New research published this week in the journal Nature examines the potential of photonic processors for artificial intelligence applications. The results demonstrate for the first time that these devices can process information quickly and in parallel, something that today’s electronic chips cannot do.

“Neural networks ‘learn’ by absorbing huge data sets and recognizing patterns through a series of algorithms,” explained Nathan Youngblood, assistant professor of electrical and computer engineering at the University of Pittsburgh’s Swanson School of Engineering and co-author . “This new processor would allow you to perform multiple calculations at the same time, using different optical wavelengths for each calculation. The challenge we wanted to address is integration: how can we do calculations using light in a scalable and efficient way?”

The fast and efficient processing that the researchers sought is ideal for applications such as autonomous vehicles, which need to process the data they detect from multiple inputs as quickly as possible. Photonic processors can also support applications in cloud computing, medical imaging and more.

“Light-based processors to accelerate tasks in the field of machine learning allow complex mathematical tasks to be processed at high speeds and throughput,” said senior co-author Wolfram Pernice, from the University of Münster. “This is much faster than conventional chips that rely on electronic data transfer, such as graphics cards or specialized hardware such as TPUs (Tensor Processing Unit).”

The research was conducted by an international team of researchers, including Pitt, the University of Münster in Germany, the Universities of Oxford and Exeter in England, the École Polytechnique Fédérale (EPFL) in Lausanne, Switzerland, and the IBM Research Laboratory in Zurich .

The researchers combined phase-shift materials – the storage material used, for example, on DVDs – and photonic structures to store data in a non-volatile manner, without requiring a continuous power supply. This study is also the first to combine these optical memory cells with a chip-based frequency comb as a light source, which allowed them to calculate at 16 different wavelengths simultaneously.

In the article, the researchers used the technology to create a convolutional neural network that would recognize handwritten numbers. They found that the method provided data rates and computing densities never seen before.

“The convolutional operation between input data and one or more filters – which can be a border highlight on a photo, for example – can be transferred very well to our matrix architecture,” said Johannes Feldmann, a graduate student at the University Münster and lead author of the study. “Exploiting light for signal transfer allows the processor to perform parallel data processing through wavelength multiplexing, which leads to higher computation density and many matrix multiplications being performed in just one step of time. In contrast to traditional electronics, which generally operate at low GHz ranges, optical modulation speeds can be achieved at speeds up to 50 to 100 GHz. ”

The article, “Parallel convolution processing using an integrated photonic tensor core”, was published in Nature and co-authored by Johannes Feldmann, Nathan Youngblood, Maxim Karpov, Helge Gehring, Xuan Li, Maik Stappers, Manuel Le Gallo, Xin Fu, Anton Lukashchuk, Arslan Raja, Junqiu Liu, David Wright, Abu, Tobias Kippenberg, Wolfram Pernice, and Harish Bhaskaran.