aiqbench analysis
This research addresses a critical stability issue in event-based vision systems by proving that hard temporal binning introduces discontinuities sensitive to minute timestamp shifts. The authors demonstrate that linear two-bin interpolation achieves optimal L1 sensitivity bounds, significantly reducing representation drift without compromising accuracy on benchmarks like DVS Gesture and N-MNIST. For practitioners building robust spiking neural networks or processing asynchronous sensor streams, adopting continuous encoders is essential to prevent prediction flips caused by minor timing variations.
On The Robustness-Resolution Tradeoff In Temporal Quantization Of Event Streams
arXiv:2609.22295v2 Announce Type: replace-cross Abstract: Event pipelines often discretize asynchronous timestamps before learning. This step looks harmless, but its stability depends directly on temporal resolution. We study this dependence at the representation level. We first show that hard temporal binning is discontinuous: an arbitrarily small timestamp shift near a boundary can move unit event mass between bins. We then define a class of nonnegative, mass-preserving, resolution-faithful continuous encoders and prove that every encoder in this class has global L1 sensitivity at least 2/Delta, where Delta denotes bin width. Linear two-bin interpolation attains this limit. Local support and first-moment preservation also make it unique. Experiments on SHD, N-MNIST, and DVS128 Gesture support the analysis. Across uniform timestamp budgets, linear interpolation lowers mean representation drift by 47-72% while keeping clean accuracy nearly unchanged. On DVS Gesture, it produces zero prediction flips across all tested budgets and three seeds. On SHD, measured drift follows 1/Delta with R^2 = 0.992.
Abstract:Event pipelines often discretize asynchronous timestamps before learning. This step looks harmless, but its stability depends directly on temporal resolution. We study this dependence at the representation level. We first show that hard temporal binning is discontinuous: an arbitrarily small timestamp shift near a boundary can move unit event mass between bins. We then define a class of nonnegative, mass-preserving, resolution-faithful continuous encoders and prove that every encoder in this class has global L1 sensitivity at least 2/Delta, where Delta denotes bin width. Linear two-bin interpolation attains this limit. Local support and first-moment preservation also make it unique. Experiments on SHD, N-MNIST, and DVS128 Gesture support the analysis. Across uniform timestamp budgets, linear interpolation lowers mean representation drift by 47-72% while keeping clean accuracy nearly unchanged. On DVS Gesture, it produces zero prediction flips across all tested budgets and three seeds. On SHD, measured drift follows 1/Delta with R^2 = 0.992.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2609.22295 [cs.CV]
(or arXiv:2609.22295v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.22295 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 249 words
Related coverage
REACT: A Fully Spiking State-Space Model for Real-Time Event-Driven Temporal Perception
arXiv · Sep 18, 2026
MotionDLO: Hybrid Event- and Frame-Based Tracking of Deformable Linear Objects
arXiv · Aug 25, 2026
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Latent Space · Sep 25, 2026
REALM: An RGB- and Event-Aligned Latent Manifold for Cross-Modal Perception
arXiv · Sep 22, 2026