{"id":"https://openalex.org/W7137177504","doi":"https://doi.org/10.48550/arxiv.2603.12354","title":"Alternating Gradient Flow Utility: A Unified Metric for Structural Pruning and Dynamic Routing in Deep Networks","display_name":"Alternating Gradient Flow Utility: A Unified Metric for Structural Pruning and Dynamic Routing in Deep Networks","publication_year":2026,"publication_date":"2026-03-12","ids":{"openalex":"https://openalex.org/W7137177504","doi":"https://doi.org/10.48550/arxiv.2603.12354"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.12354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.12354","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.12354","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128669770","display_name":"Tianhao Qian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qian, Tianhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037460817","display_name":"Zhuoxuan Li","orcid":"https://orcid.org/0000-0002-2096-6414"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhuoxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129547074","display_name":"Jinde Cao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cao, Jinde","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045991552","display_name":"Xinli Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Xinli","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Rutkowski, Leszek","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rutkowski, Leszek","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.37700000405311584,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.37700000405311584,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.219200000166893,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.11159999668598175,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.4542999863624573},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.439300000667572},{"id":"https://openalex.org/keywords/traverse","display_name":"Traverse","score":0.41819998621940613},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.39879998564720154},{"id":"https://openalex.org/keywords/rendering","display_name":"Rendering (computer graphics)","score":0.38769999146461487},{"id":"https://openalex.org/keywords/heuristics","display_name":"Heuristics","score":0.3743000030517578},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.3702999949455261},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.3684000074863434}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7263000011444092},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.4542999863624573},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4408999979496002},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.439300000667572},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4242999851703644},{"id":"https://openalex.org/C176809094","wikidata":"https://www.wikidata.org/wiki/Q15401496","display_name":"Traverse","level":2,"score":0.41819998621940613},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.39879998564720154},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.38769999146461487},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.3743000030517578},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3702999949455261},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.3684000074863434},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.319599986076355},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.31139999628067017},{"id":"https://openalex.org/C199845137","wikidata":"https://www.wikidata.org/wiki/Q145490","display_name":"Network topology","level":2,"score":0.3005000054836273},{"id":"https://openalex.org/C184720557","wikidata":"https://www.wikidata.org/wiki/Q7825049","display_name":"Topology (electrical circuits)","level":2,"score":0.29840001463890076},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.29510000348091125},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.28630000352859497},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2800999879837036},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C114809511","wikidata":"https://www.wikidata.org/wiki/Q1412924","display_name":"Flow network","level":2,"score":0.26080000400543213},{"id":"https://openalex.org/C108154423","wikidata":"https://www.wikidata.org/wiki/Q1469792","display_name":"Salience (neuroscience)","level":2,"score":0.2581000030040741},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.12354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.12354","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.12354","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.12354","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Efficient":[0],"deep":[1,36],"learning":[2],"traditionally":[3],"relies":[4],"on":[5,183,192,215],"static":[6],"heuristics":[7],"like":[8],"weight":[9],"magnitude":[10,45],"or":[11],"activation":[12],"awareness":[13],"(e.g.,":[14],"Wanda,":[15],"RIA).":[16],"While":[17],"successful":[18],"in":[19,107,127,145],"unstructured":[20],"settings,":[21],"we":[22,56,83,120,136,160],"observe":[23],"a":[24,44,58,85,122,132,162,187],"critical":[25,50],"limitation":[26],"when":[27,209],"applying":[28],"these":[29,157],"metrics":[30,41,202],"to":[31,48,74,114],"the":[32,77,104,197,225,228,243],"structural":[33,79,118,170,198],"pruning":[34],"of":[35,124,227,239],"vision":[37],"networks.":[38],"These":[39],"contemporary":[40],"suffer":[42,141],"from":[43,110,142,172],"bias,":[46],"failing":[47],"preserve":[49],"functional":[51],"pathways.":[52],"To":[53],"overcome":[54],"this,":[55],"propose":[57],"decoupled":[59],"kinetic":[60],"paradigm":[61,182],"inspired":[62],"by":[63,156,231],"Alternating":[64],"Gradient":[65],"Flow":[66],"(AGF),":[67],"utilizing":[68],"an":[69,235],"absolute":[70],"feature-space":[71],"Taylor":[72],"expansion":[73],"accurately":[75],"capture":[76],"network's":[78],"\"kinetic":[80],"utility\".":[81],"First,":[82],"uncover":[84],"topological":[86,100],"phase":[87],"transition":[88],"at":[89],"extreme":[90],"sparsity,":[91],"where":[92,200],"AGF":[93,194],"successfully":[94],"preserves":[95],"baseline":[96],"functionality":[97],"and":[98],"exhibits":[99],"implicit":[101],"regularization,":[102],"avoiding":[103],"collapse":[105,199],"seen":[106],"models":[108],"trained":[109],"scratch.":[111],"Second,":[112],"transitioning":[113],"architectures":[115],"without":[116,241],"strict":[117],"priors,":[119],"reveal":[121],"phenomenon":[123],"Sparsity":[125],"Bottleneck":[126],"Vision":[128],"Transformers":[129],"(ViTs).":[130],"Through":[131],"gradient-magnitude":[133],"decoupling":[134],"analysis,":[135],"discover":[137],"that":[138,166],"dynamic":[139,213],"signals":[140],"signal":[143],"compression":[144,189],"converged":[146],"models,":[147],"rendering":[148],"them":[149],"suboptimal":[150],"for":[151,212],"real-time":[152],"routing.":[153],"Finally,":[154],"driven":[155],"empirical":[158],"constraints,":[159],"design":[161],"hybrid":[163,218],"routing":[164],"framework":[165],"decouples":[167],"AGF-guided":[168],"offline":[169],"search":[171],"online":[173],"execution":[174],"via":[175],"zero-cost":[176],"physical":[177],"priors.":[178],"We":[179],"validate":[180],"our":[181,217],"large-scale":[184],"benchmarks:":[185],"under":[186],"75%":[188],"stress":[190],"test":[191],"ImageNet-1K,":[193],"effectively":[195],"avoids":[196],"traditional":[201],"aggressively":[203],"fall":[204],"below":[205],"random":[206],"sampling.":[207],"Furthermore,":[208],"systematically":[210],"deployed":[211],"inference":[214],"ImageNet-100,":[216],"approach":[219],"achieves":[220],"Pareto-optimal":[221],"efficiency.":[222],"It":[223],"reduces":[224],"usage":[226],"heavy":[229],"expert":[230],"approximately":[232],"50%":[233],"(achieving":[234],"estimated":[236],"overall":[237],"cost":[238],"0.92$\\times$)":[240],"sacrificing":[242],"full-model":[244],"accuracy.":[245]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-17T00:00:00"}