Noisy models of the ventral stream reveal the impact of recurrence and learned representations on information processing timescales

{"In":[0],"vision":[1],"neuroscience,":[2],"the":[3,7,22,25,37,76,83,116,120,155,161,165,176,184,187,193,198,204,207,215,218,227,230,234,241,250],"temporal":[4,117,162,231],"dynamics":[5],"of":[6,11,24,27,56,109,119,123,154,164,175,186,197,201,206,217,229,233,249],"sensory":[8,77],"stream":[9],"and":[10,50,114,172,245],"its":[12],"neural":[13,89,107,166,235],"representations":[14,42,108,125],"are":[15,44,59],"thought":[16],"to":[17,21,61,70,214],"be":[18],"deeply":[19],"linked":[20],"function":[23,246],"hierarchy":[26],"cortical":[28,105],"areas":[29],"that":[30,43,68,98],"deal":[31],"with":[32],"object":[33,48,57],"recognition,":[34],"known":[35],"as":[36,100,237],"visual":[38,178,252],"ventral":[39,156,177],"stream.":[40,157,179],"Neural":[41],"invariant":[45],"under":[46],"identity-preserving":[47],"transformations,":[49],"therefore":[51],"allow":[52],"for":[53,240],"efficient":[54],"learning":[55,66],"identity,":[58],"theorized":[60],"emerge":[62],"from":[63,75],"a":[64,169,238],"self-supervised":[65],"process":[67],"attempts":[69],"extract":[71],"\\"temporally":[72],"stable\\"":[73],"features":[74,196],"input.":[78],"Conversely,":[79],"invariance":[80],"increases":[81],"along":[82,103],"hierarchy,":[84,106],"putatively":[85],"implying":[86],"progressively":[87],"slower":[88],"codes":[90,167],"in":[91,139,151,168,247],"higher-level":[92],"areas.":[93],"Recent":[94],"neurophysiological":[95],"evidence":[96],"shows":[97],"indeed,":[99],"one":[101],"moves":[102],"this":[104],"dynamic":[110],"stimuli":[111],"become":[112],"slower,":[113],"additionally":[115],"scales":[118],"within-trial":[121],"fluctuations":[122],"these":[124,133],"(called":[126],"\\"intrinsic":[127],"timescales\\")":[128],"increase":[129],"starkly.":[130],"However,":[131],"while":[132,203],"timescale":[134],"hierarchies":[135],"have":[136,147],"been":[137],"reproduced":[138],"biologically":[140],"grounded":[141],"recurrent":[142,171],"models,":[143],"their":[144],"network":[145],"determinants":[146],"remained":[148],"largely":[149],"unexplored":[150],"image-computable":[152],"models":[153,248],"Here":[158],"we":[159],"investigate":[160],"structure":[163,232,244],"noisy,":[170],"adaptive":[173],"model":[174],"We":[180],"show":[181],"that,":[182],"surprisingly,":[183],"organization":[185],"representation":[188],"timescales":[189,209],"is":[190,212],"set":[191],"by":[192,221],"broad":[194],"architectural":[195],"network,":[199],"regardless":[200],"training,":[202],"ordering":[205],"intrinsic":[208],"across":[210],"layers":[211],"sensitive":[213],"details":[216],"functions":[219],"implemented":[220],"each":[222],"layer.":[223],"Our":[224],"work":[225],"underscores":[226],"importance":[228],"code":[236],"probe":[239],"link":[242],"between":[243],"vertebrate":[251],"system.":[253]}

Authors

Institutions

Publication Details

Journal
PLoS Computational Biology
Published
2026-08-27
DOI
https://doi.org/10.1371/journal.pcbi.1014653
Primary Topic
Visual perception and processing mechanisms
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
article

Noisy models of the ventral stream reveal the impact of recurrence and learned representations on information processing timescales

Sara Varetti, Sebastian Goldt, Eugenio Piasini
PLoS Computational Biology
Visual perception and processing mechanisms
article

Noisy models of the ventral stream reveal the impact of recurrence and learned representations on information processing timescales

Sara Varetti, Sebastian Goldt, Eugenio Piasini
article en

Abstract

In vision neuroscience, the temporal dynamics of the sensory stream and of its neural representations are thought to be deeply linked to the function of the hierarchy of cortical areas that deal with object recognition, known as the visual ventral stream. Neural representations that are invariant under identity-preserving object transformations, and therefore allow for efficient learning of object identity, are theorized to emerge from a self-supervised learning process that attempts to extract "temporally stable" features from the sensory input. Conversely, invariance increases along the hierarchy, putatively implying progressively slower neural codes in higher-level areas. Recent neurophysiological evidence shows that indeed, as one moves along this cortical hierarchy, neural representations of dynamic stimuli become slower, and additionally the temporal scales of the within-trial fluctuations of these representations (called "intrinsic timescales") increase starkly. However, while these timescale hierarchies have been reproduced in biologically grounded recurrent models, their network determinants have remained largely unexplored in image-computable models of the ventral stream. Here we investigate the temporal structure of the neural codes in a noisy, recurrent and adaptive model of the ventral visual stream. We show that, surprisingly, the organization of the representation timescales is set by the broad architectural features of the network, regardless of training, while the ordering of the intrinsic timescales across layers is sensitive to the details of the functions implemented by each layer. Our work underscores the importance of the temporal structure of the neural code as a probe for the link between structure and function in models of the vertebrate visual system.

PLoS Computational BiologyVol. 22(8)
Scuola Internazionale Superiore di Studi Avanzati (IT)
European Commission
Quality Education
Openalex Percentile: Top 90%
Visual perception and processing mechanisms
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.