[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85587-en":3,"doc-seo-85587-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85587,1649267921044,"Ava Thompson","https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1782875107921204101",8,"Research & Report","Congestion-Aware Dynamic Axonal Delay for Spiking Neural Networks","Spiking Neural Networks (SNNs) provide an energy-efficient framework for processing temporal, event-driven information, where spike timing alignment is crucial for representational power and training stability. Existing delay-learning methods typically use static synapse delays, leading to many parameters and weak adaptation to input-dependent activity dynamics. A Congestion-Aware Dynamic Axonal Delay (CADAD) mechanism decomposes delays into channel-wise static bases and a global activity-conditioned dynamic shift, learned via differentiable interpolation and discretized at inference. Experiments on SHD, SSC, and GSC-35 improve temporal-task accuracy and reduce delay parameters by about 50%.","arXiv :2605 .0 129 1v 3 [ cs .LG] 13 Jul 2026  \nCongestion-Aware Dynamic Axonal Delay for Spiking  \nNeural Networks  \nDewei Bai* Hongxiang Peng* Yunyun Zeng Ziyu Zhang  \nHong Qu\\#  \nUniversity of Electronic Science and Technology of China  \n*Equal contribution. \\# Corresponding author.  \nAbstract  \nSpiking Neural Networks (SNNs) are widely regarded as an energy-efficient paradigm for modeling and processing temporal and event-driven information.  \nIncorporating delays in SNNs has been proven to be an effective mechanism for improving spike alignment in event-driven tasks. However, existing delay learning approaches predominantly assign static delays to individual synapses, resulting ina large number of delay parameters and limited adaptability to input-dependent activity dynamics. To this end, we propose a Congestion-Aware Dynamic Axonal Delay (CADAD) mechanism, which decomposes the delay into a channel-wise static base delay for temporal structuring and a global, activity-conditioned shift that dynamically regulates the state update rate under varying spike intensities.  \nThe delay parameters are learned using differentiable linear interpolation and discretized at inference time, preserving the benefits of dynamic delay modulation while incurring only minimal additional cost. Experiments on speech benchmarks, including the Spiking Heidelberg Dataset, Spiking Speech Commands, and Google Speech Commands, demonstrate that introducing congestion-aware delays into synaptic signal transmission effectively improves accuracy on temporal tasks, notably achieving 93.75% accuracy on SHD, 80.69% accuracy on SSC, and 95.58% on GSC-35, while reducing the parameter count by approximately 50% compared to state-of-the-art delay-based methods with the same architecture.  \n1 Introduction  \nOwing to their event-driven computational paradigm, Spiking Neural Networks (SNNs) exhibit significant advantages in low-power and energy-efficient intelligent computing. In such networks, information propagates across space and time via discrete spikes, with the representational capacity and training stability being highly dependent on the temporal alignment, propagation, and integration of spike signals.  \nFrom the perspective of computational neuroscience, spiking neurons can be regarded as timesensitive coincidence detectors König et al. [1996], Rossant et al. [2011], where effective information processing relies on the relative timing of spike arrivals rather than the firing rate of a single neuron. Signal delays along axons and synapses directly shape these arrival times, and the heterogeneity of such delays provides critical degrees of freedom for modeling complex spatiotemporal patterns Izhikevich [2006] . Extensive biological research further indicates that synaptic and transmission delays are not static; instead, they can be regulated through learning processes, suggesting that temporal regulation itself is a fundamental component of neural computation Bowers [2017] .  \nInspired by these findings, recent studies have introduced learnable transmission delays into SNNs to explicitly enhance temporal modeling capabilities, achieving performance gains across various time  \nPreprint.  \n\n|  |  |  |\n| --- | --- | --- |\n|  |  |  |\n\nt1  \nt2  \nTime  \n(a) No delay  \n\n|  | d1 d2 |  |  |  | Overflow |\n| --- | --- | --- | --- | --- | --- |\n|  |  |  |  |  |  |\n|  |  |  |  |  |  |\n|  |  |  |  |  |  |\n\nVth  \nt1  \nt1+d1t2 t2+d2  \nTime  \nMembranne Pooootential  \n(b) Static delay  \n\n|  | d[t1] d[t2] |  |  |  | Overflow |\n| --- | --- | --- | --- | --- | --- |\n|  |  |  |  |  |  |\n|  |  |  |  |  |  |\n\nVth  \nt1+d[t1] t2 t2+d[t2] Time  \nMembrbn Potentiantia  \n(c) Dynamic delay  \nFigure 1: Conceptual comparison of delay mechanisms. (a) Without delays, spikes arrive at t1 and t2 asynchronously, causing information loss due to membrane leakage. (b) Static delays add fixed time intervals, improving integration but resulting in misalignment and broad wasteful overflow (red arrow","cbCaibCyAZFxA6wQ","https://ap.wps.com/l/cbCaibCyAZFxA6wQ","pdf",2711629,1,20,"English","en",105,"# Abstract\n# Introduction\n## Motivation: spike timing alignment and limitations of static delays\n## Proposed method: Congestion-Aware Dynamic Axonal Delay (CADAD)","[{\"question\":\"What problem do static-delay learning approaches face in spiking neural networks?\",\"answer\":\"They assign static delays per synapse, which creates many delay parameters and limited adaptability to input-dependent, non-stationary spike activity.\"},{\"question\":\"How does CADAD model axonal delays differently from existing methods?\",\"answer\":\"CADAD decomposes delays into channel-wise static base delays for temporal structure and a global activity-conditioned dynamic shift that regulates state update rates under varying spike intensities.\"},{\"question\":\"How are the delay parameters trained and used during inference?\",\"answer\":\"They are learned using differentiable linear interpolation and then discretized at inference time, preserving dynamic modulation with minimal additional cost.\"}]",1784204763,50,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"congestion-aware-dynamic-axonal-delay-for-spiking-neural-networks","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/congestion-aware-dynamic-axonal-delay-for-spiking-neural-networks/85587/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem do static-delay learning approaches face in spiking neural networks?","Question",{"text":75,"@type":76},"They assign static delays per synapse, which creates many delay parameters and limited adaptability to input-dependent, non-stationary spike activity.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does CADAD model axonal delays differently from existing methods?",{"text":80,"@type":76},"CADAD decomposes delays into channel-wise static base delays for temporal structure and a global activity-conditioned dynamic shift that regulates state update rates under varying spike intensities.",{"name":82,"@type":73,"acceptedAnswer":83},"How are the delay parameters trained and used during inference?",{"text":84,"@type":76},"They are learned using differentiable linear interpolation and then discretized at inference time, preserving dynamic modulation with minimal additional cost.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":28,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]