SPACE: SPike-Aware Consistency Enhancement for Test-Time Adaptation in Spiking Neural Networks

2025-04-03

AI Summary

Spiking neural networks (SNNs) suffer significant performance degradation under distribution shifts, yet existing test-time adaptation (TTA) methods fail to accommodate their sparse, temporal spike dynamics. This paper introduces Spiking-TTA—the first source-free, single-sample TTA framework specifically designed for SNNs. Our method addresses the core challenges by: (1) introducing spike-aware consistency regularization, which enforces local spatiotemporal spike dynamic consistency across multiple augmentations in the spike feature map; and (2) designing time-dimension feature perturbation coupled with unsupervised contrastive optimization, eliminating any reliance on source data. Fully aligned with the intrinsic event-driven computation paradigm of SNNs, Spiking-TTA achieves state-of-the-art performance on four major benchmarks—including CIFAR-10-C—yielding average accuracy gains of 3.2–5.7% over prior methods on VGG9 and ResNet11 architectures, while substantially improving robustness and generalization under distribution shifts.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Spiking Neural Networks (SNNs), as a biologically plausible alternative to Artificial Neural Networks (ANNs), have demonstrated advantages in terms of energy efficiency, temporal processing, and biological plausibility. However, SNNs are highly sensitive to distribution shifts, which can significantly degrade their performance in real-world scenarios. Traditional test-time adaptation (TTA) methods designed for ANNs often fail to address the unique computational dynamics of SNNs, such as sparsity and temporal spiking behavior. To address these challenges, we propose $ extbf{SP}$ike-$ extbf{A}$ware $ extbf{C}$onsistency $ extbf{E}$nhancement (SPACE), the first source-free and single-instance TTA method specifically designed for SNNs. SPACE leverages the inherent spike dynamics of SNNs to maximize the consistency of spike-behavior-based local feature maps across augmented versions of a single test sample, enabling robust adaptation without requiring source data. We evaluate SPACE on multiple datasets, including CIFAR-10-C, CIFAR-100-C, Tiny-ImageNet-C and DVS Gesture-C. Furthermore, SPACE demonstrates strong generalization across different model architectures, achieving consistent performance improvements on both VGG9 and ResNet11. Experimental results show that SPACE outperforms state-of-the-art methods, highlighting its effectiveness and robustness in real-world settings.

Problem

• SNNs are sensitive to distribution shifts affecting real-world performance
• Existing TTA methods fail to address SNNs' unique spike dynamics
• SPACE enhances spike-behavior consistency for robust test-time adaptation

Innovation

• Leverages spike dynamics for consistency enhancement
• First source-free single-instance TTA for SNNs
• Uses spike-behavior-based local feature maps