Devices called photonic tensor processing units rely on light to perform matrix-multiplication operations for machine-learning algorithms. Dong et al. propose a computer architecture that uses low-quality light from LEDs for efficient energy consumption.
AI algorithms are energy-intensive, spurring the need for more energy-efficient methods. Replacing parts of digital computers with analogue components, like photonic tensor processing units, powered by LEDs, can address this challenge.
[
Collection
]
[
|
...
]