Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Functional integration and inference in the brain.

Karl Friston1

  • 1The Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, 12 Queen Square, London, UK. k.friston@fil.ion.ucl.ac.uk

Progress in Neurobiology
|November 27, 2002
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Scale-invariant evolution: Comment on "homo informatio" by Michael Walker.

Physics of life reviews·2026
Same author

The body does not keep the score: trauma, predictive coding, and the restoration of metastability.

Frontiers in systems neuroscience·2026
Same author

The dysconnection hypothesis of schizophrenia: a 30-year update.

World psychiatry : official journal of the World Psychiatric Association (WPA)·2026
Same author

The methodological foundations of lesion network mapping remain sound.

bioRxiv : the preprint server for biology·2026
Same author

Insula Structure Is Linked to Autonomic Cardiac Dysregulation in Depression.

Biological psychiatry·2026
Same author

Correction: Dynamic causal models in infectious disease epidemiology-an assessment of their predictive validity based on the COVID-19 epidemic in the UK 2020 to 2024.

Frontiers in public health·2026
Same journal

Deletion of Fbxo25 causes excessive repetitive behavior, impaired recognition memory, reduced dendritic complexity, and aberrant protein expression in mice.

Progress in neurobiology·2026
Same journal

Neuropsychiatric Disease Mechanisms and Interventions. from 22q11.2 Deletion Syndrome Experimental Studies.

Progress in neurobiology·2026
Same journal

REM sleep as a dummy-model of the world: A theoretical framework.

Progress in neurobiology·2026
Same journal

CA3 transiently modulates spatial representation in CA1.

Progress in neurobiology·2026
Same journal

Love, death, and oxytocin: In memory of Larry Young.

Progress in neurobiology·2026
Same journal

Assessing peripheral oxytocin and cortisol levels and epigenetic variations of oxytocin receptor and glucocorticoid receptor genes in school-aged preterm-born children.

Progress in neurobiology·2026
See all related articles

This review explores brain models for sensory input, differentiating between redundancy reduction and prediction error minimization. It highlights the necessity of backward connections for generative models in sensory processing and learning.

Area of Science:

  • Neuroscience
  • Computational Neuroscience
  • Machine Learning

Background:

  • Self-supervised models of brain's sensory processing fall into two categories: minimizing redundancy or prediction error.
  • These models, while sharing goals, differ in implementation and functional architecture.
  • Existing models often assume prior knowledge of sensory cause distributions.

Purpose of the Study:

  • To review and compare two classes of self-supervised brain models for sensory input representation.
  • To explore the role of backward connections in generative models for perceptual learning.
  • To investigate how empirical evidence can distinguish between different functional architectures.

Main Methods:

  • Review of existing literature on self-supervised models in neuroscience.

Related Experiment Videos

  • Application of empirical Bayes to demonstrate learning priors in a hierarchical context.
  • Analysis of generative models and the necessity of backward connections for non-invertible processes.
  • Discussion of functional integration using effective connectivity and functional neuroimaging.

    • Main Results:

    • Backward connections are essential for generative models when input-generating processes are non-invertible.
    • Feedforward architectures alone are insufficient for perceptual learning and synthesis.
    • Learned priors in a hierarchical context eliminate the need for pre-specified assumptions.
    • Modulatory backward connections enable interaction between higher and lower cortical levels.

    Conclusions:

    • Generative models, mediated by backward connections, are crucial for sensory processing.
    • Functional neuroimaging can reveal interactions between bottom-up and top-down processing.
    • Empirical evidence supports the prevalence of top-down influences and the plausibility of generative models in sensory brain function.