ElfCore: A 28nm Neural Processor Enabling Dynamic Structured Sparse Training and Online Self-Supervised Learning with Activity-Dependent Weight Update
This work addresses the problem of high power and memory demands in neural processors for real-time sensory applications, offering a significant improvement over existing methods.
The paper tackles the challenge of efficient event-driven sensory signal processing by introducing ElfCore, a 28nm neural processor that integrates online self-supervised learning, dynamic structured sparse training, and activity-dependent weight updates, resulting in up to 16x lower power consumption, 3.8x reduced memory, and 5.9x greater network capacity efficiency compared to state-of-the-art solutions.
In this paper, we present ElfCore, a 28nm digital spiking neural network processor tailored for event-driven sensory signal processing. ElfCore is the first to efficiently integrate: (1) a local online self-supervised learning engine that enables multi-layer temporal learning without labeled inputs; (2) a dynamic structured sparse training engine that supports high-accuracy sparse-to-sparse learning; and (3) an activity-dependent sparse weight update mechanism that selectively updates weights based solely on input activity and network dynamics. Demonstrated on tasks including gesture recognition, speech, and biomedical signal processing, ElfCore outperforms state-of-the-art solutions with up to 16X lower power consumption, 3.8X reduced on-chip memory requirements, and 5.9X greater network capacity efficiency.