Online Continual Learning on Intel Loihi 2 via a Co-designed Spiking Neural Network

2025/11/03 by Elvin Hajizada, Hajizada, Elvin, Danielle Rager +13
Computer Science · Engineering · #Advanced Memory and Neural Computing #Advanced Neural Network Applications #Artificial Intelligence (cs.AI) #Distributed #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Neural and Evolutionary Computing (cs.NE) #Parallel #and Cluster Computing (cs.DC)

paper · pdf · doi:10.48550/arxiv.2511.01553

openalex publication_date 2025/11/03 · openalex created_date 2025/11/06 · openalex updated_date 2026/07/28

Abstract

AI systems on edge devices require online continual learning -- adapting to non-stationary streams and unfamiliar classes without catastrophic forgetting -- under strict power constraints. We present CLP-SNN, a spiking neural network with a self-normalizing local learning rule and a spike-driven neural state machine for autonomous on-chip learning, implemented on Intel's Loihi 2 neuromorphic processor. On OpenLORIS few-shot experiments, CLP-SNN matches replay-based accuracy rehearsal-free. On Loihi 2, CLP-SNN achieves 113x lower latency (0.33 ms vs. 37.3 ms) and 6,600x lower energy (0.05 mJ vs. 333 mJ) than the strongest edge-GPU baseline. This gain decomposes into algorithmic efficiency (~14.5x latency, ~22.6x energy on the same GPU) and neuromorphic hardware co-design (~7.8x latency, ~295x energy) exploiting event-driven learning and sparse graded-spike communication. We show that co-designed brain-inspired algorithms and neuromorphic hardware can break traditional accuracy-efficiency trade-offs in edge AI.

Citations

• Online Continual Learning: A Systematic Literature Review of Approaches, Challenges, and Benchmarks
• Continual Learning for Autonomous Robots: A Prototype-based Approach
• Training Machine Learning models at the Edge: A Survey
• Efficient Rehearsal Free Zero Forgetting Continual Learning using Adaptive Weight Modulation
• Efficient Video and Audio processing with Loihi 2
• Neurogenesis Dynamics-inspired Spiking Neural Network Training Acceleration
• On the Stability-Plasticity Dilemma of Class-Incremental Learning
• A Comprehensive Survey of Continual Learning: Theory, Method and Application
• On-Device Training Under 256KB Memory
• Online Continual Learning for Embedded Devices
• Efficient Neuromorphic Signal Processing with Loihi 2
• Continual Learning for Real-World Autonomous Systems: Algorithms, Challenges and Frameworks
• Continual Learning at the Edge: Real-Time Training on Smartphone Devices
• Advancing Neuromorphic Computing With Loihi: A Survey of Results and Outlook
• TinyOL: TinyML with Online-Learning on Microcontrollers
• Towards Open World Object Detection
• Online Continual Learning in Image Classification: An Empirical Survey
• A Review of Open-World Learning and Steps Toward Open-World Learning Without Labels
• Online Few-shot Gesture Learning on a Neuromorphic Processor
• Wandering Within a World: Online Contextualized Few-Shot Learning
• Measuring Robustness to Natural Distribution Shifts in Image Classification
• Language Models are Few-Shot Learners
• OpenLORIS-Object: A Robotic Vision Dataset and Benchmark for Lifelong Deep Learning
• EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
• Training on the Edge: The why and the how
• Simplified neuron model as a principal component analyzer

Cited by

• A flexible framework for structural plasticity in GPU-accelerated sparse spiking neural networks

Related

• Continual Learning with Neuromorphic Computing: Theories, Methods, and Applications
• Replay4NCL: An Efficient Memory Replay-based Methodology for Neuromorphic Continual Learning in Embedded AI Systems
• Genesis: A Spiking Neuromorphic Accelerator With On-chip Continual Learning
• Emulating Brain-like Rapid Learning in Neuromorphic Edge Computing
• Continual Learning for Autonomous Robots: A Prototype-based Approach