{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,27]],"date-time":"2026-02-27T06:24:25Z","timestamp":1772173465593,"version":"3.50.1"},"update-to":[{"DOI":"10.1371\/journal.pcbi.1012822","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,2,27]],"date-time":"2025-02-27T00:00:00Z","timestamp":1740614400000}}],"reference-count":41,"publisher":"Public Library of Science (PLoS)","issue":"2","license":[{"start":{"date-parts":[[2025,2,19]],"date-time":"2025-02-19T00:00:00Z","timestamp":1739923200000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.ploscompbiol.org"],"crossmark-restriction":false},"short-container-title":["PLoS Comput Biol"],"abstract":"<jats:p>Voxelwise encoding models based on convolutional neural networks (CNNs) are widely used as predictive models of brain activity evoked by natural movies. Despite their superior predictive performance, the huge number of parameters in CNN-based models have made them difficult to interpret. Here, we investigate whether model compression can build more interpretable and more stable CNN-based voxelwise models while maintaining accuracy. We used multiple compression techniques to prune less important CNN filters and connections, a receptive field compression method to select receptive fields with optimal center and size, and principal component analysis to reduce dimensionality. We demonstrate that the model compression improves the accuracy of identifying visual stimuli in a hold-out test set. Additionally, compressed models offer a more stable interpretation of voxelwise pattern selectivity than uncompressed models. Finally, the receptive field-compressed models reveal that the optimal model-based population receptive fields become larger and more centralized along the ventral visual pathway. Overall, our findings support using model compression to build more interpretable voxelwise models.<\/jats:p>","DOI":"10.1371\/journal.pcbi.1012822","type":"journal-article","created":{"date-parts":[[2025,2,19]],"date-time":"2025-02-19T13:32:42Z","timestamp":1739971962000},"page":"e1012822","update-policy":"https:\/\/doi.org\/10.1371\/journal.pcbi.corrections_policy","source":"Crossref","is-referenced-by-count":2,"title":["Compression-enabled interpretability of voxelwise encoding models"],"prefix":"10.1371","volume":"21","author":[{"given":"Fatemeh","family":"Kamali","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6842-3099","authenticated-orcid":true,"given":"Amir Abolfazl","family":"Suratgar","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mohammadbagher","family":"Menhaj","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7824-4628","authenticated-orcid":true,"given":"Reza","family":"Abbasi-Asl","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"340","published-online":{"date-parts":[[2025,2,19]]},"reference":[{"issue":"7185","key":"pcbi.1012822.ref001","doi-asserted-by":"crossref","first-page":"352","DOI":"10.1038\/nature06713","article-title":"Identifying natural images from human brain activity","volume":"452","author":"KN Kay","year":"2008","journal-title":"Nature"},{"issue":"2","key":"pcbi.1012822.ref002","doi-asserted-by":"crossref","first-page":"400","DOI":"10.1016\/j.neuroimage.2010.07.073","article-title":"Encoding and decoding in fMRI","volume":"56","author":"T Naselaris","year":"2011","journal-title":"Neuroimage"},{"key":"pcbi.1012822.ref003","article-title":"Pixels to voxels: modeling visual representation in the human brain.","author":"P Agrawal","year":"2014"},{"key":"pcbi.1012822.ref004","article-title":"Neural encoding and interpretation for high-level visual cortices based on fmri using image caption features.","author":"K Qiao","year":"2020"},{"key":"pcbi.1012822.ref005","doi-asserted-by":"crossref","first-page":"125","DOI":"10.1016\/j.neuroimage.2019.05.039","article-title":"Variational autoencoder: an unsupervised model for encoding and decoding fMRI activity in visual cortex","volume":"198","author":"K Han","year":"2019","journal-title":"Neuroimage"},{"issue":"6","key":"pcbi.1012822.ref006","doi-asserted-by":"crossref","first-page":"902","DOI":"10.1016\/j.neuron.2009.09.006","article-title":"Bayesian reconstruction of natural images from human brain activity","volume":"63","author":"T Naselaris","year":"2009","journal-title":"Neuron"},{"key":"pcbi.1012822.ref007","doi-asserted-by":"crossref","first-page":"482","DOI":"10.1016\/j.neuroimage.2019.04.012","article-title":"Voxelwise encoding models with non-spherical multivariate normal priors","volume":"197","author":"AO Nunez-Elizalde","year":"2019","journal-title":"Neuroimage"},{"issue":"12","key":"pcbi.1012822.ref008","doi-asserted-by":"crossref","first-page":"4136","DOI":"10.1093\/cercor\/bhx268","article-title":"Neural encoding and decoding with deep learning for dynamic natural vision","volume":"28","author":"H Wen","year":"2018","journal-title":"Cereb Cortex"},{"issue":"12","key":"pcbi.1012822.ref009","doi-asserted-by":"crossref","first-page":"2037","DOI":"10.1109\/TPAMI.2006.244","article-title":"Face description with local binary patterns: application to face recognition","volume":"28","author":"T Ahonen","year":"2006","journal-title":"IEEE Trans Pattern Anal Mach Intell"},{"key":"pcbi.1012822.ref010","doi-asserted-by":"crossref","first-page":"7","DOI":"10.3389\/fncom.2017.00007","article-title":"Modeling the dynamics of human brain activity with recurrent neural networks","volume":"11","author":"U G\u00fc\u00e7l\u00fc","year":"2017","journal-title":"Front Comput Neurosci"},{"issue":"05","key":"pcbi.1012822.ref011","doi-asserted-by":"crossref","first-page":"2140017","DOI":"10.1142\/S0219519421400170","article-title":"Dense-gwp: an improved primary visual encoding model based on dense gabor features","volume":"21","author":"Y Cui","year":"2021","journal-title":"J Mech Med Biol"},{"issue":"11","key":"pcbi.1012822.ref012","doi-asserted-by":"crossref","first-page":"e1003915","DOI":"10.1371\/journal.pcbi.1003915","article-title":"Deep supervised, but not unsupervised, models may explain IT cortical representation","volume":"10","author":"S-M Khaligh-Razavi","year":"2014","journal-title":"PLoS Comput Biol"},{"issue":"27","key":"pcbi.1012822.ref013","doi-asserted-by":"crossref","first-page":"10005","DOI":"10.1523\/JNEUROSCI.5023-14.2015","article-title":"Deep neural networks reveal a gradient in the complexity of neural representations across the ventral stream","volume":"35","author":"U G\u00fc\u00e7l\u00fc","year":"2015","journal-title":"J Neurosci"},{"issue":"1","key":"pcbi.1012822.ref014","doi-asserted-by":"crossref","first-page":"329","DOI":"10.1016\/j.neuroimage.2015.12.036","article-title":"Increasingly complex representations of natural movies across the dorsal stream are shared between subjects","volume":"145","author":"U G\u00fc\u00e7l\u00fc","year":"2017","journal-title":"Neuroimage"},{"key":"pcbi.1012822.ref015","doi-asserted-by":"crossref","first-page":"184","DOI":"10.1016\/j.neuroimage.2016.10.001","article-title":"Seeing it all: convolutional network layers map the function of the human visual system","volume":"152","author":"M Eickenberg","year":"2017","journal-title":"Neuroimage"},{"issue":"Pt A","key":"pcbi.1012822.ref016","doi-asserted-by":"crossref","first-page":"188","DOI":"10.1016\/j.neuroimage.2017.06.035","article-title":"The feature-weighted receptive field: an interpretable encoding model for complex feature spaces","volume":"180","author":"G St-Yves","year":"2018","journal-title":"Neuroimage"},{"key":"pcbi.1012822.ref017","doi-asserted-by":"crossref","first-page":"253","DOI":"10.1016\/j.neuroimage.2017.07.018","article-title":"Convolutional neural network-based encoding and decoding of visual object recognition in space and time","volume":"180","author":"K Seeliger","year":"2018","journal-title":"NeuroImage"},{"issue":"1","key":"pcbi.1012822.ref018","first-page":"5408942","article-title":"A comparative analysis of visual encoding models based on classification and segmentation task\u2010driven CNNs","author":"Z Yu","year":"2020","journal-title":"Comput Math Methods Med"},{"issue":"5","key":"pcbi.1012822.ref019","doi-asserted-by":"crossref","first-page":"2269","DOI":"10.1002\/hbm.24006","article-title":"Deep recurrent neural network reveals a hierarchy of process memory during dynamic natural vision","volume":"39","author":"J Shi","year":"2018","journal-title":"Hum Brain Mapp"},{"issue":"1","key":"pcbi.1012822.ref020","doi-asserted-by":"crossref","first-page":"3752","DOI":"10.1038\/s41598-018-22160-9","article-title":"Deep residual network predicts cortical representation and organization of visual features for rapid categorization","volume":"8","author":"H Wen","year":"2018","journal-title":"Sci Rep"},{"key":"pcbi.1012822.ref021","article-title":"Accurate reconstruction of image stimuli from human fMRI based on the decoding model with capsule network architecture.","author":"K Qiao","year":"2018"},{"issue":"44","key":"pcbi.1012822.ref022","doi-asserted-by":"crossref","first-page":"22071","DOI":"10.1073\/pnas.1900654116","article-title":"Definitions, methods, and applications in interpretable machine learning","volume":"116","author":"W Murdoch","year":"2019","journal-title":"Proc Natl Acad Sci"},{"key":"pcbi.1012822.ref023","article-title":"Towards a rigorous science of interpretable machine learning.","author":"F Doshi-Velez","year":"2017"},{"key":"pcbi.1012822.ref024","article-title":"Multi-modal prototype learning for interpretable multivariable time series classification.","author":"G Ghosal","year":"2021"},{"key":"pcbi.1012822.ref025","doi-asserted-by":"crossref","first-page":"704182","DOI":"10.3389\/fdata.2021.704182","article-title":"Structural compression of convolutional neural networks with applications in interpretability","volume":"4","author":"R Abbasi-Asl","year":"2021","journal-title":"Front Big Data"},{"key":"pcbi.1012822.ref026","unstructured":"Abbasi-Asl R, Yu B. Interpreting convolutional neural networks through compression. arXiv:1711.02329 [Preprint]. 2017."},{"key":"pcbi.1012822.ref027","article-title":"Deep compression: compressing deep neural networks with pruning, trained quantization and huffman coding.","author":"S Han","year":"2015"},{"issue":"4","key":"pcbi.1012822.ref028","doi-asserted-by":"crossref","first-page":"873","DOI":"10.1109\/TSMCB.2006.870645","article-title":"A novel incremental principal component analysis and its application for face recognition","volume":"36","author":"H Zhao","year":"2006","journal-title":"IEEE Tran Syst Man Cybern Part B (Cybern)"},{"key":"pcbi.1012822.ref029","doi-asserted-by":"crossref","first-page":"105","DOI":"10.1016\/j.neuroimage.2013.04.127","article-title":"The minimal preprocessing pipelines for the human connectome project","volume":"80","author":"M Glasser","year":"2013","journal-title":"NeuroImage"},{"issue":"7615","key":"pcbi.1012822.ref030","doi-asserted-by":"crossref","first-page":"171","DOI":"10.1038\/nature18933","article-title":"A multi-modal parcellation of human cerebral cortex","volume":"536","author":"M Glasser","year":"2016","journal-title":"Nature"},{"key":"pcbi.1012822.ref031","author":"S Nishimoto","year":"2014"},{"issue":"1","key":"pcbi.1012822.ref032","doi-asserted-by":"crossref","first-page":"233","DOI":"10.1016\/j.neuroimage.2004.05.012","article-title":"Parametric reverse correlation reveals spatial linearity of retinotopic human V1 BOLD response","volume":"23","author":"K Hansen","year":"2004","journal-title":"Neuroimage"},{"issue":"19","key":"pcbi.1012822.ref033","doi-asserted-by":"crossref","first-page":"1641","DOI":"10.1016\/j.cub.2011.08.031","article-title":"Reconstructing visual experiences from brain activity evoked by natural movies","volume":"21","author":"S Nishimoto","year":"2011","journal-title":"Curr Biol"},{"key":"pcbi.1012822.ref034","article-title":"Imagenet classification with deep convolutional neural networks","volume":"25","author":"A Krizhevsky","year":"2012","journal-title":"Adv Neural Inf Process Syst"},{"issue":"1","key":"pcbi.1012822.ref035","doi-asserted-by":"crossref","first-page":"93","DOI":"10.1002\/wics.14","article-title":"Ridge regression","volume":"1","author":"G McDonald","year":"2009","journal-title":"Wiley Interdiscip Rev Comput Stat"},{"key":"pcbi.1012822.ref036","doi-asserted-by":"crossref","DOI":"10.1016\/B978-012372560-8\/50014-0","article-title":"Convolution models for fMRI.","volume-title":"Statistical parametric mapping: the analysis of functional brain images","author":"R Henson","year":"2007"},{"key":"pcbi.1012822.ref037","first-page":"465534","article-title":"The DeepTune framework for modeling and characterizing neurons in visual cortex area V4","author":"R Abbasi-Asl","year":"2018","journal-title":"bioRxiv"},{"issue":"1","key":"pcbi.1012822.ref038","doi-asserted-by":"crossref","first-page":"106","DOI":"10.1113\/jphysiol.1962.sp006837","article-title":"Receptive fields, binocular interaction and functional architecture in the cat\u2019s visual cortex","volume":"160","author":"DH Hubel","year":"1962","journal-title":"J Physiol"},{"issue":"2","key":"pcbi.1012822.ref039","doi-asserted-by":"crossref","first-page":"647","DOI":"10.1016\/j.neuroimage.2007.09.034","article-title":"Population receptive field estimates in human visual cortex","volume":"39","author":"SO Dumoulin","year":"2008","journal-title":"Neuroimage"},{"key":"pcbi.1012822.ref040","doi-asserted-by":"crossref","first-page":"e67304","DOI":"10.7554\/eLife.67304","article-title":"Population receptive fields in nonhuman primates from whole-brain fMRI and large-scale neurophysiology in visual cortex","volume":"10","author":"P Klink","year":"2021","journal-title":"Elife"},{"issue":"1","key":"pcbi.1012822.ref041","doi-asserted-by":"crossref","first-page":"116","DOI":"10.1038\/s41593-021-00962-x","article-title":"A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence","volume":"25","author":"E Allen","year":"2021","journal-title":"Nat Neurosci"}],"updated-by":[{"DOI":"10.1371\/journal.pcbi.1012822","type":"new_version","label":"New version","source":"publisher","updated":{"date-parts":[[2025,2,27]],"date-time":"2025-02-27T00:00:00Z","timestamp":1740614400000}}],"container-title":["PLOS Computational Biology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012822","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,2,27]],"date-time":"2025-02-27T13:37:45Z","timestamp":1740663465000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pcbi.1012822"}},"subtitle":[],"editor":[{"given":"Daniele","family":"Marinazzo","sequence":"first","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2025,2,19]]},"references-count":41,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2025,2,19]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pcbi.1012822","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2022.07.09.494042","asserted-by":"object"}]},"ISSN":["1553-7358"],"issn-type":[{"value":"1553-7358","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,2,19]]}}}