aiqbench analysis
This research introduces A2SG, a unified framework that resolves training challenges in deep spiking neural networks by adapting gradient windows and leveraging asymmetric dynamics to improve convergence and generalization. By establishing a theoretical link between gradient variation and loss landscape curvature, the method consistently enhances accuracy and energy efficiency across CNN and Transformer architectures. The approach offers a robust solution for deploying SNNs on neuromorphic hardware, with code already released for community adoption.
A2SG:Adaptive and Asymmetric Surrogate Gradients for Training Deep Spiking Neural Networks
arXiv:2606.11236v1 Announce Type: cross Abstract: Training deep spiking neural networks (SNNs) remains challenging due to sharp loss landscapes and temporal inconsistency caused by surrogate gradients. To address these challenges, we propose a unified framework: adaptive and asymmetric surrogate gradients A2SG. The adaptive gradients adjust an effective window for spatio-temporal adaptation, reducing spatial gradient variation and maintaining directional consistency of gradients over time. The asymmetric gradients reflect neuronal dynamics by assigning larger gradients to neurons with higher membrane potentials, and we prove that they yield lower variation than symmetric surrogates. Our analysis further establishes a direct connection between local gradient variation and the curvature of the loss landscape, providing a principled explanation for how A2SG promotes convergence to flatter minima and improves generalization. We conduct extensive experiments on diverse models, including CNN-based and Transformer-based SNNs, across various tasks such as image classification using both static and neuromorphic datasets, as well as segmentation. The results demonstrate that A2SG consistently improves accuracy and energy efficiency, establishing it as a general and reliable solution for training deep SNNs. Our code is available at https://github.com/KIST-NCL/A2SG.git.
Abstract:Training deep spiking neural networks (SNNs) remains challenging due to sharp loss landscapes and temporal inconsistency caused by surrogate gradients. To address these challenges, we propose a unified framework: adaptive and asymmetric surrogate gradients A2SG. The adaptive gradients adjust an effective window for spatio-temporal adaptation, reducing spatial gradient variation and maintaining directional consistency of gradients over time. The asymmetric gradients reflect neuronal dynamics by assigning larger gradients to neurons with higher membrane potentials, and we prove that they yield lower variation than symmetric surrogates. Our analysis further establishes a direct connection between local gradient variation and the curvature of the loss landscape, providing a principled explanation for how A2SG promotes convergence to flatter minima and improves generalization. We conduct extensive experiments on diverse models, including CNN-based and Transformer-based SNNs, across various tasks such as image classification using both static and neuromorphic datasets, as well as segmentation. The results demonstrate that A2SG consistently improves accuracy and energy efficiency, establishing it as a general and reliable solution for training deep SNNs. Our code is available at this https URL.
Neural and Evolutionary Computing (cs.NE); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2606.11236 [cs.NE]
(or arXiv:2606.11236v1 [cs.NE] for this version)
Full text under the publisher's open licence · 199 words
Related coverage
Adversarial Prompts for Acceptance Collapse in Speculative Decoding
arXiv · Jul 27, 2026
CARM: Conformal Adaptive Risk Masking for Reliable Scientific Super-Resolution
arXiv · Nov 5, 2026
DDSNet: Dual-domain Symmetry-aware Network for PCSEL Property Prediction
arXiv · Jul 29, 2026
ACPO: Asymmetric Credit Policy Optimization via Mode-Local Entropy Surrogate
arXiv · Jul 31, 2026