New paper demonstrates use of photonic structures for AI


Machine Learning at the Speed ​​of Light: New Paper Demonstrates Use of Photonic Structures for AI

Illustration showing parallel convolutional processing using an integrated phonetic tensor nucleus. New research published this week in the journal Nature examines the potential of photonic processors for artificial intelligence applications. Credit: XVIVO

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 calculation methods are not up to the task, but in searching for a solution, the researchers saw the light – literally.

Light-based processors, called photonic processors, allow computers to perform 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 today’s microchips cannot.

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

The fast and efficient processing researchers are looking for 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, etc.

“Light-based processors for speeding up tasks in machine learning make it possible to process complex mathematical tasks at high speeds and rates,” said Wolfram Pernice, co-lead author of the University of Münster . “It’s much faster than conventional chips that rely on electronic data transfer, such as graphics cards or specialized hardware like TPUs (Tensor Processing Unit).”

The research was carried out 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.

Light-based processors improve machine learning processing

Schematic representation of a processor for matrix multiplication operating in light. Credit: University of Oxford

The researchers combined phase change materials – the storage material used, for example, on DVDs – and photonic structures to store data in a non-volatile manner without requiring a continuous energy supply. This study is also the first to combine these optical memory cells with a smart frequency comb as the light source, which allowed them to calculate on 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 offered unprecedented data rates and computational densities.

“The convolutional operation between the input data and one or more filters – which can be a highlight of the edges of a photo, for example – can be transferred very well to our matrix architecture”, said Johannes Feldmann, graduate student at the University of Münster and lead author of the study. “Harnessing light for signal transfer allows the processor to perform parallel data processing by wavelength multiplexing, which leads to higher computational density and many matrix multiplications performed in only one time step. Unlike traditional electronics, which typically operate in the low GHz, optical modulation speeds can be achieved with speeds up to 50 to 100 GHz. ”

The article, “Parallel convolution processing using an integrated photonic tensor nucleus”, was published in Nature and co-written 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 Sebastian, Tobias Kippenberg, Wolfram Pernice , and Harish Bhaskaran.

Photon-based processing units enable more complex machine learning

More information:
J. Feldmann et al. Parallel convolution processing using an integrated photonic tensor nucleus, Nature (2021). DOI: 10.1038 / s41586-020-03070-1

Provided by the University of Pittsburgh

Quote: Machine Learning at the Speed ​​of Light: New Article Demonstrates Use of Photonic Structures for AI (2021, January 6) Retrieved January 7, 2021 from 01-machine-paper-photonic-ai .html

This document is subject to copyright. Apart from any fair use for study or private research, no part may be reproduced without written permission. The content is provided for information only.

Source link