RRAM & Stacked Memory: Boosting AI Performance with In-Memory Computing

Researchers at the University of California, San Diego have achieved a breakthrough in resistive random-access memory (RRAM) technology, redesigning its operational principles to potentially accelerate the processing speed of artificial neural networks.

The team, led by electrical engineer Duygu Kuzum, focused on fundamentally altering how RRAM switches states, a change that could unlock new possibilities for localized artificial intelligence applications. Kuzum, a professor in the Electrical and Computer Engineering department at UCSD and a Joan and Irwin Jacobs-Kavli Foundation Chancellor Endowed Faculty Fellow, explained the core of the innovation: “We actually redesigned RRAM, completely rethinking the way it switches,” according to reporting from MSN.

Kuzum’s research broadly centers on applying advancements in nanoelectronics to better understand computation within the brain. Her work includes the development of nanoelectronic synaptic devices that mimic the synaptic computation and plasticity observed in the human brain, with the goal of creating portable and energy-efficient computers capable of real-time learning and information processing. She also investigates neural interfaces for studying brain circuits, and is developing tools for high-precision brain probing, as detailed on the Jacobs School of Engineering website.

The redesigned RRAM approach aims to address limitations that have hindered the widespread adoption of the technology, despite its potential. TechSpot reported that while RRAM has yet to fully deliver on its promises, the UCSD team’s work represents a significant step toward realizing its potential for stacked memory applications.

Kuzum received her Ph.D. In electrical engineering from Stanford University in 2010 and completed a postdoctoral fellowship in bioengineering at the University of Pennsylvania from 2011 to 2015. Her work has been featured in publications including Nano Letters, Nature, New Scientist, and EE Times, and she has authored or co-authored over 40 journal and conference papers, according to her UCSD faculty profile.

According to Google Scholar, Kuzum’s research has garnered over 10,771 citations, and she collaborates with researchers including H.-S. Philip Wong of Stanford University and Krishna Saraswat, also of Stanford. The next steps for the technology and its potential impact on the field remain to be seen.