NeuSpin: Design of a Reliable Edge Neuromorphic System Based on Spintronics for Green AI

TL;DR

This work addresses the hardware hurdles of deploying Bayesian Neural Networks at the edge by proposing NeuSpin, a full-stack co-design that leverages spintronic computation-in-memory (CIM) architectures. It develops dropout- and variational-inference–based BayNNs tailored for spintronic memories, introducing SpinDrop, MC-SpatialDropout, SpinScaleDropout, Inverted Normalization with Affine Dropout, Bayesian Sub-Set Parameter Inference, and SpinBayes to efficiently realize uncertainty estimation with CIM. Key contributions include dramatic energy and memory reductions (up to ~$100 imes$ and ~ $158.7 imes$, respectively), robust out-of-distribution detection (up to 100%), and improved inference on corrupted data, achieved through algorithm-hardware co-design and device-aware training. The results demonstrate that BayNNs can be deployed at the edge with reliable uncertainty quantification, offering significant practical impact for Green AI in wearable and IoT health applications, with validation across spintronic MTJ/CIM hardware concepts and realistic data.

Abstract

Internet of Things (IoT) and smart wearable devices for personalized healthcare will require storing and computing ever-increasing amounts of data. The key requirements for these devices are ultra-low-power, high-processing capabilities, autonomy at low cost, as well as reliability and accuracy to enable Green AI at the edge. Artificial Intelligence (AI) models, especially Bayesian Neural Networks (BayNNs) are resource-intensive and face challenges with traditional computing architectures due to the memory wall problem. Computing-in-Memory (CIM) with emerging resistive memories offers a solution by combining memory blocks and computing units for higher efficiency and lower power consumption. However, implementing BayNNs on CIM hardware, particularly with spintronic technologies, presents technical challenges due to variability and manufacturing defects. The NeuSPIN project aims to address these challenges through full-stack hardware and software co-design, developing novel algorithmic and circuit design approaches to enhance the performance, energy-efficiency and robustness of BayNNs on sprintronic-based CIM platforms.

Figures (3)

• Figure 1: Crossbar design for the MC-SpatialDropout based on mapping strategy (a) ① and (b) strategy②. In (b), only the Dropout module and WL decoder are shown. Everything else is abstracted.
• Figure 2: Proposed inference architecture for Scale-Dropout.
• Figure 3: Proposed layer architecture