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 Video

Updated: Mar 6, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

1.2K

Fixed versus mixed RSA: Explaining visual representations by fixed and mixed feature sets from shallow and deep

Seyed-Mahdi Khaligh-Razavi1, Linda Henriksson2, Kendrick Kay3

  • 1MRC Cognition and Brain Sciences Unit, Cambridge, UK; Computer Science & Artificial intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, MA, USA.

Journal of Mathematical Psychology
|March 17, 2017
PubMed
Summary

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

Objective quality assessment for precision functional MRI data.

Neuron·2026
Same author

Integrating neuroscience across species and scales.

Nature neuroscience·2026
Same author

A Systematic Characterization of Causal Interactions Between Human Visual Areas.

bioRxiv : the preprint server for biology·2026
Same author

Dissociating stimulus encoding and task demands in ECoG responses from human visual cortex.

bioRxiv : the preprint server for biology·2026
Same author

The colors of images preferred by individual voxels can be used to delineate functionally distinct visually responsive brain areas.

Proceedings of the National Academy of Sciences of the United States of America·2026
Same author

Objective Quality Assessment for Precision Functional MRI Data.

bioRxiv : the preprint server for biology·2026

Representational Similarity Analysis (RSA) compares computational models to brain activity. Shallow models fit early visual areas, while deep convolutional networks with feature mixing are essential for higher visual processing.

Area of Science:

  • Neuroscience
  • Computational Vision
  • Machine Learning

Background:

  • Computational models of object vision are increasingly used to understand the primate visual system.
  • Fitting complex models to limited brain-activity data is challenging.
  • Representational Similarity Analysis (RSA) offers a method to compare model and brain representations.

Purpose of the Study:

  • To investigate the effectiveness of different computational models in explaining visual cortex representations.
  • To compare shallow and deep models using Representational Similarity Analysis (RSA).
  • To determine the necessity of linear feature transformation (mixing) for model fitting.

Main Methods:

  • Functional magnetic resonance imaging (fMRI) was used to measure brain responses to natural images.
Keywords:
Deep convolutional networksMixed RSAObject-vision modelsRepresentational similarity analysisVoxel-receptive-field modelling

Related Experiment Videos

Last Updated: Mar 6, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

1.2K
  • Representational Similarity Analysis (RSA) was applied to compare model-derived representational dissimilarity matrices with brain data.
  • Both fixed and mixed RSA approaches were used, with mixed RSA involving linear transformation of model features.

    • Main Results:

    • Early visual areas were well-explained by shallow models like the Gabor wavelet pyramid (GWP), with no significant benefit from feature mixing.
    • Higher visual areas in the ventral stream (lateral occipital region) were best explained by deep convolutional networks.
    • Linear transformation (mixing) of deep convolutional network features was essential for explaining the representational space in higher visual areas.

    Conclusions:

    • Shallow models adequately capture representations in early visual cortex.
    • Deep convolutional networks, particularly with feature mixing, are crucial for understanding higher-level visual processing.
    • The training regime of deep networks may influence their representational structure and the necessity of feature mixing for explaining brain data.