In a recent edition of Advanced Photonics Research, University of Utah electrical and computer engineering assistant professor, Weilu Gao presented a new optoelectronic architecture for accelerating machine learning hardware. He believes his findings will provide an efficient solution to the issue of carbon emissions, especially for large data centers.

“Data centers consume a lot of power and contribute a lot to the carbon footprint, so reducing their power consumption is very important. The photonic optics approach can be a way of solving this issue,” Gao said.

Gao has also been invited to host a poster presentation on the topic at the renowned Conference on Laser and Electro-Optics this May. This year’s event will be held virtually. He is excited to use this opportunity to share his research with the community and is confident that his work will lay the foundation for future research that will ultimately bring his architecture to life.

“When we have some acknowledgment from the community,” Gao said, “then we can continue our work to push forward to the real implementation of this hardware.”

In the publication, Gao outlines a framework that not only addresses the issue of non-uniformity in nanostructures that currently limits large-scale hardware development but also requires ultralow energy consumption while still providing high data throughput.

“The detailed analysis together with the proof-of-concept algorithm demonstration shows that the proposed architecture is promising for a high-throughput and power-efficient platform for accelerating machine-learning algorithms,” Gao said.

His framework relies on two key components: large arrays of high-performing graphene spatial light modulators and tunable responsivity photodetectors.

In addition to the normal challenges that come with performing research, Gao’s team has been greatly impacted by the COVID-19 pandemic. It has been especially hard on his students, whose work primarily requires in-person, hands-on efforts. Gao has had to remain flexible throughout the process and has relied on the cooperation of his colleagues.

Despite the obstacles, he is looking forward to more collaborations and research with faculty in the future.

Read Gao’s full journal publication here. His co-authors include ECE assistant professor Cunxi Yu and student Ruiyang Chen.