Year: 2025
Type: article
Abstract: Custom integrated circuits modeling biological neural networks serve as tools for studying brain computation and platforms for exploring new architectures and learning rules of artificial neural networks. Time synchronization across network units is an important aspect of these designs to ensure reproducible results and maintain hardware-software equivalence. Current approaches rely on global synchronization protocols, which fundamentally limit system scalability. To overcome this, we develop NeuroScale, a decentralized and scalable neuromorphic architecture that uses local, aperiodic synchronization to preserve determinism without global coordination. Cores of co-localized compute and memory elements model neural and synaptic processes, including spike filtering operations, subthreshold neural dynamics, and online Hebbian learning rules. Multiple cores communicate via spikes across a routing mesh, using distributed event-driven synchronization to efficiently scale to large networks. We compare this synchronization protocol to the global barrier synchronization approaches of IBM TrueNorth and Intel Loihi, demonstrating NeuroScale's advantages for large system sizes.
Source: Nature Communications
Authors: Congyang Li, Nabil Imam, Rajit Manohar
Institutions: Yale University, Georgia Institute of Technology
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $6,990
Awards: NCS-FR: Insect-based brain-machine interfaces and robots for understanding odor-driven navigation, G2702044909, THIS COOPERATIVE AGREEMENT SHALL BE PERFORMED IN ACCORDANCE WITH STATEMENT OF WORK, ENTITLED 'IDEAL: AN INTELLIGENT DESIGN ENVIRONMENT FOR ASYNCHRONO, EFRI BRAID: Rapid contextual learning in resilient autonomous systems
OpenAlex ID: W4416594646