Hybrid Neuromorphic-Deep Learning Systems for AI Acceleration in Edge Computing

Hybrid Neuromorphic-Deep Learning for Edge AI Acceleration

Abstract

The growing demand for energy-efficient and responsive artificial intelligence (AI) systems at the edge has intensified interest in neuromorphic computing, which mimics the brain’s mechanisms to enable low-power, real-time data processing. While neuromorphic systems excel in energy efficiency, their scalability and broader applicability remain constrained. To address these limitations, this study introduces a hybrid framework that combines spiking neural networks (SNNs) with conventional deep learning architectures such as convolutional neural networks (CNNs). By leveraging the strengths of both paradigms, the proposed system enhances AI acceleration for edge computing environments characterized by resource constraints. A detailed mathematical representation of the hybrid system is developed, followed by performance
evaluations using established datasets. The results highlight significant gains in energy efficiency, achieving reductions of up to 35%, alongside latency improvements of up to 45% compared to existing neuromorphic and traditional AI methods. Moreover, the system demonstrates scalability and adaptability to diverse edge applications, including Internet of Things (IoT) devices and autonomous systems. These findings underline the transformative potential of hybrid neuromorphic-deep learning architectures in advancing the capabilities of next-generation edge AI while bridging the gap between bioinspired and conventional computational methods.