Jointly efficient encoding and decoding in neural populations

The efficient coding approach proposes that neural systems represent as much sensory information as biological constraints allow. It aims at formalizing encoding as a constrained optimal process. A different approach, that aims at formalizing decoding, proposes that neural systems instantiate a gene...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:bioRxiv
Hlavní autoři: Malerba, Simone Blanco, Micheli, Aurora, Woodford, Michael, da Silveira, Rava Azeredo
Médium: Paper
Jazyk:angličtina
Vydáno: Cold Spring Harbor Laboratory 04.05.2024
Vydání:1.2
Témata:
Neuroscience
ISSN:2692-8205
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract The efficient coding approach proposes that neural systems represent as much sensory information as biological constraints allow. It aims at formalizing encoding as a constrained optimal process. A different approach, that aims at formalizing decoding, proposes that neural systems instantiate a generative model of the sensory world. Here, we put forth a normative framework that characterizes neural systems as jointly optimizing encoding and decoding. It takes the form of a variational autoencoder: sensory stimuli are encoded in the noisy activity of neurons to be interpreted by a flexible decoder; encoding must allow for an accurate stimulus reconstruction from neural activity. Jointly, neural activity is required to represent the statistics of latent features which are mapped by the decoder into distributions over sensory stimuli; decoding correspondingly optimizes the accuracy of the generative model. This framework yields in a family of encoding-decoding models, which result in equally accurate generative models, indexed by a measure of the stimulus-induced deviation of neural activity from the marginal distribution over neural activity. Each member of this family predicts a specific relation between properties of the sensory neurons—such as the arrangement of the tuning curve means (preferred stimuli) and widths (degrees of selectivity) in the population—as a function of the statistics of the sensory world. Our approach thus generalizes the efficient coding approach. Notably, here, the form of the constraint on the optimization derives from the requirement of an accurate generative model, while it is arbitrary in efficient coding models. Moreover, solutions do not require the knowledge of the stimulus distribution, but are learned on the basis of data samples; the constraint further acts as regularizer, allowing the model to generalize beyond the training data. Finally, we characterize the family of models we obtain through alternate measures of performance, such as the error in stimulus reconstruction. We find that a range of models admits comparable performance; in particular, a population of sensory neurons with broad tuning curves as observed experimentally yields both low reconstruction stimulus error and an accurate generative model that generalizes robustly to unseen data. Our brain represents the sensory world in the activity of populations of neurons. Two theories have addressed the nature of these representations. The first theory—efficient coding—posits that neurons encode as much information as possible about sensory stimuli, subject to resource constraints such as limits on energy consumption. The second one—generative modeling—focuses on decoding, and is organized around the idea that neural activity plays the role of a latent variable from which sensory stimuli can be simulated. Our work subsumes the two approaches in a unifying framework based on the mathematics of variational autoencoders. Unlike in efficient coding, which assumes full knowledge of stimulus statistics, here representations are learned from examples, in a joint optimization of encoding and decoding. This new framework yields a range of optimal representations, corresponding to different models of neural selectivity and reconstruction performances, depending on the resource constraint. The form of the constraint is not arbitrary but derives from the optimization framework, and its strength tunes the ability of the model to generalize beyond the training example. Central to the approach, and to the nature of the representations it implies, is the interplay of encoding and decoding, itself central to brain processing.
AbstractList The efficient coding approach proposes that neural systems represent as much sensory information as biological constraints allow. It aims at formalizing encoding as a constrained optimal process. A different approach, that aims at formalizing decoding, proposes that neural systems instantiate a generative model of the sensory world. Here, we put forth a normative framework that characterizes neural systems as jointly optimizing encoding and decoding. It takes the form of a variational autoencoder: sensory stimuli are encoded in the noisy activity of neurons to be interpreted by a flexible decoder; encoding must allow for an accurate stimulus reconstruction from neural activity. Jointly, neural activity is required to represent the statistics of latent features which are mapped by the decoder into distributions over sensory stimuli; decoding correspondingly optimizes the accuracy of the generative model. This framework yields in a family of encoding-decoding models, which result in equally accurate generative models, indexed by a measure of the stimulus-induced deviation of neural activity from the marginal distribution over neural activity. Each member of this family predicts a specific relation between properties of the sensory neurons—such as the arrangement of the tuning curve means (preferred stimuli) and widths (degrees of selectivity) in the population—as a function of the statistics of the sensory world. Our approach thus generalizes the efficient coding approach. Notably, here, the form of the constraint on the optimization derives from the requirement of an accurate generative model, while it is arbitrary in efficient coding models. Moreover, solutions do not require the knowledge of the stimulus distribution, but are learned on the basis of data samples; the constraint further acts as regularizer, allowing the model to generalize beyond the training data. Finally, we characterize the family of models we obtain through alternate measures of performance, such as the error in stimulus reconstruction. We find that a range of models admits comparable performance; in particular, a population of sensory neurons with broad tuning curves as observed experimentally yields both low reconstruction stimulus error and an accurate generative model that generalizes robustly to unseen data. Our brain represents the sensory world in the activity of populations of neurons. Two theories have addressed the nature of these representations. The first theory—efficient coding—posits that neurons encode as much information as possible about sensory stimuli, subject to resource constraints such as limits on energy consumption. The second one—generative modeling—focuses on decoding, and is organized around the idea that neural activity plays the role of a latent variable from which sensory stimuli can be simulated. Our work subsumes the two approaches in a unifying framework based on the mathematics of variational autoencoders. Unlike in efficient coding, which assumes full knowledge of stimulus statistics, here representations are learned from examples, in a joint optimization of encoding and decoding. This new framework yields a range of optimal representations, corresponding to different models of neural selectivity and reconstruction performances, depending on the resource constraint. The form of the constraint is not arbitrary but derives from the optimization framework, and its strength tunes the ability of the model to generalize beyond the training example. Central to the approach, and to the nature of the representations it implies, is the interplay of encoding and decoding, itself central to brain processing.
Author Woodford, Michael
Malerba, Simone Blanco
da Silveira, Rava Azeredo
Micheli, Aurora
Author_xml – sequence: 1
  givenname: Simone Blanco
  orcidid: 0000-0002-4467-5988
  surname: Malerba
  fullname: Malerba, Simone Blanco
  email: simone.bmalerba@gmail.com
  organization: Institute for Neural Information Processing, Center for Molecular Neurobiology, University Medical Center Hamburg-Eppendorf
– sequence: 2
  givenname: Aurora
  surname: Micheli
  fullname: Micheli, Aurora
  organization: Laboratoire de Physique de l’Ecole Normale Supérieure, ENS, Université PSL, CNRS,3 Sorbonne Université, Université de Paris
– sequence: 3
  givenname: Michael
  surname: Woodford
  fullname: Woodford, Michael
  organization: Department of Economics, Columbia University
– sequence: 4
  givenname: Rava Azeredo
  surname: da Silveira
  fullname: da Silveira, Rava Azeredo
  organization: Faculty of Science, University of Basel
BookMark eNotT01LxDAUDKLguu4P8Jajl9aXr7fJwYMsfrLgRc8lbV4lUpPSdsX991Z3mcMwzDDMXLDTlBMxdiWgFALEjQSpSsBSitJoY7U9YQuJThZWgjlnq3H8BADpUKi1XrDblxzT1O05tW1sIqWJU2pyiOmD-xR4oKOIiSfaDb7jfe53nZ9iTuMlO2t9N9LqyEv2_nD_tnkqtq-Pz5u7bVHPm2xhTa2BZBNw3ZI0jXA4IyirwXmPNZI3rhZ2jljjkPDfFmpeqRRaUkt2feitYx5-4nfVD_HLD_vq720FWElRHd6qX8LqSt8
Cites_doi 10.1162/neco.1995.7.5.889
10.1038/s41467-022-35659-7
10.1073/pnas.1918386117/-/DCSupplemental.y
10.1162/089976699300016809
10.1038/s41467-021-26751-5
10.1101/2022.12.12.519538
10.1016/j.neuron.2016.09.038
10.1162/NECOa00804
10.1515/znc-1981-9-1040
10.1101/399246
10.1038/nn.2901
10.48550/arXiv.1312.6114
10.1007/BF00188924
10.1126/science.1195870
10.1007/BF00962720
10.48550/arXiv.1602.02282
10.1109/TPAMI.2017.2784440
10.1371/journal.pcbi.1008146
10.1038/nn.4105
10.1073/pnas.1004906107/-/DCSupplemental
10.1038/360068a0
10.1038/s41562-023-01584-y
10.1162/089976698300017115
10.1101/2020.05.29.124453
10.1523/JNEUROSCI.1032-14.2014
10.1162/089976698300017818
10.48550/arXiv.1401.4082
10.1101/2022.01.06.475186
10.1038/nature23020
10.48550/arXiv.1609.02200
10.1016/j.cobeha.2019.03.004
10.1371/journal.pcbi.1007476
10.48550/arXiv.1412.6980
10.1038/nn1790
10.1097/00006324-192711000-00011
10.1146/annurev-neuro-120320-082744
10.1038/nature04701
10.1162/neco.2008.03-07-486
10.1073/pnas.90.22.10749
10.7551/mitpress/9780262518420.003.0013
10.1038/236
10.1162/08997660260293247
10.1038/nature02768
10.1038/s41593-023-01458-6
10.1162/neco.1990.2.3.308
10.48550/arXiv.1906.00331
10.1162/NECOa00638
10.1007/s10957-009-9522-7
10.1111/j.2517-6161.1983.tb01251.x
10.1101/178418
10.1038/nn831
10.1364/JOSAA.20.001434
10.1038/s41593-019-0554-5
10.1097/00004647-200110000-00001
10.1038/381607a0
10.1016/j.neuron.2016.03.020
10.1561/2200000056
ContentType Paper
Copyright 2024, Posted by Cold Spring Harbor Laboratory
Copyright_xml – notice: 2024, Posted by Cold Spring Harbor Laboratory
DBID FX.
DOI 10.1101/2023.06.21.545848
DatabaseName bioRxiv
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
Discipline Biology
EISSN 2692-8205
Edition 1.2
ExternalDocumentID 2023.06.21.545848v2
GroupedDBID 8FE
8FH
AFKRA
ALMA_UNASSIGNED_HOLDINGS
BBNVY
BENPR
BHPHI
FX.
HCIFZ
LK8
M7P
NQS
PIMPY
PROAC
RHI
ID FETCH-LOGICAL-b1108-85b40e2cd67fe25c196969d38409aa6b6ea59b18e2c8596e66969d132963368e3
IngestDate Tue Jan 07 18:51:52 EST 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed false
IsScholarly false
Language English
License This pre-print is available under a Creative Commons License (Attribution-NonCommercial-NoDerivs 4.0 International), CC BY-NC-ND 4.0, as described at http://creativecommons.org/licenses/by-nc-nd/4.0
LinkModel OpenURL
MergedId FETCHMERGED-LOGICAL-b1108-85b40e2cd67fe25c196969d38409aa6b6ea59b18e2c8596e66969d132963368e3
Notes Competing Interest Statement: The authors have declared no competing interest.
ORCID 0000-0002-4467-5988
OpenAccessLink https://www.biorxiv.org/content/10.1101/2023.06.21.545848
PageCount 49
ParticipantIDs biorxiv_primary_2023_06_21_545848
PublicationCentury 2000
PublicationDate 20240504
PublicationDateYYYYMMDD 2024-05-04
PublicationDate_xml – month: 5
  year: 2024
  text: 20240504
  day: 4
PublicationDecade 2020
PublicationTitle bioRxiv
PublicationYear 2024
Publisher Cold Spring Harbor Laboratory
Publisher_xml – name: Cold Spring Harbor Laboratory
References Gjorgjieva, Meister, Sompolinsky (2023.06.21.545848v2.35) 2019; 15
Barlow (2023.06.21.545848v2.1) 1961; 1
Hahn, Wei (2023.06.21.545848v2.60) 2022
Hoyer, Hyvärinen (2023.06.21.545848v2.22) 2003; 15
Wei, Stocker (2023.06.21.545848v2.58) 2016; 28
Csikor, Meszéna, Szabó, Orbán (2023.06.21.545848v2.18) 2022
Schaffner, Bao, Tobler, Hare, Polania (2023.06.21.545848v2.30) 2023; 7
Dayan, Abbott (2023.06.21.545848v2.15) 2001
Sharpee, Berkowitz (2023.06.21.545848v2.45) 2019; 29
Jang, Gu, Poole (2023.06.21.545848v2.82) 2016
Orbán, Berkes, Fiser, Lengyel (2023.06.21.545848v2.17) 2016; 92
Lewicki (2023.06.21.545848v2.6) 2002; 5
Attwell, Laughlin (2023.06.21.545848v2.9) 2001; 21
Sønderby, Raiko, Maaløe, Sønderby, Winther (2023.06.21.545848v2.46) 2016; 30
Ma, Beck, Latham, Pouget (2023.06.21.545848v2.73) 2006; 9
Zemel, Dayan, Pouget (2023.06.21.545848v2.20) 1998; 10
Achille, Soatto (2023.06.21.545848v2.50) 2018; 40
Bozkurt, Esmaeili, Tristan, Brooks, Dy, van de Meent (2023.06.21.545848v2.70) 2021; 130
Arrow, Azawa, Hurwicz, Uzawa, Chenery, Johnson (2023.06.21.545848v2.80) 1958; 2
Tkačik, Prentice, Balasubramanian, Schneidman (2023.06.21.545848v2.37) 2010; 107
Kingma, Ba (2023.06.21.545848v2.48) 2015; 3
Barello, Charles, Pillow (2023.06.21.545848v2.26) 2018
Wei, Stocker (2023.06.21.545848v2.33) 2012; 7552
Laughlin (2023.06.21.545848v2.2) 1981; 36
Haefner, Berkes, Fiser (2023.06.21.545848v2.72) 2016; 90
van Hateren (2023.06.21.545848v2.5) 1992; 360
Benjamin, Zhang, Qiu, Stocker, Kording (2023.06.21.545848v2.27) 2022; 13
Wang, Stocker, Lee (2023.06.21.545848v2.29) 2012; 25
Ganguli, Simoncelli (2023.06.21.545848v2.57) 2016
Aridor, Grechi, Woodford (2023.06.21.545848v2.67) 2020
Ganguli, Simoncelli (2023.06.21.545848v2.10) 2014; 26
Zhaoping (2023.06.21.545848v2.7) 2014
Higgins, Chang, Langston, Hassabis, Summerfield, Tsao (2023.06.21.545848v2.66) 2021; 12
Boyd, Boyd, Vandenberghe (2023.06.21.545848v2.78) 2004
Rezende, Mohamed, Wierstra (2023.06.21.545848v2.24) 2014; 31
Bethge, Rotermund, Pawelzik (2023.06.21.545848v2.51) 2002; 14
Lin, Jin, Jordan (2023.06.21.545848v2.79) 2020; 37
Dayan, Hinton, Neal, Zemel (2023.06.21.545848v2.14) 1995; 7
Von Helmholtz (2023.06.21.545848v2.13) 1927; 4
van Hateren (2023.06.21.545848v2.4) 1992; 171
Wainwright, Simoncelli (2023.06.21.545848v2.16) 1999; 12
Yerxa, Kee, DeWeese, Cooper (2023.06.21.545848v2.56) 2020; 16
Rolfe (2023.06.21.545848v2.83) 2016
Walker, Cotton, Ma, Tolias (2023.06.21.545848v2.39) 2020; 23
Kingma, Welling (2023.06.21.545848v2.23) 2014; 2
Azeredo da Silveira, Rieke (2023.06.21.545848v2.42) 2021; 44
Moore (2023.06.21.545848v2.61) 1973; 54
Zhao, Ren, Yuan, Song, Goodman, Ermon (2023.06.21.545848v2.69) 2018; 32
Nedić, Ozdaglar (2023.06.21.545848v2.77) 2009; 142
Higgins, Matthey, Pal, Burgess, Glorot, Botvinick (2023.06.21.545848v2.47) 2017; 5
Rozell, Johnson, Baraniuk, Olshausen (2023.06.21.545848v2.63) 2008; 20
Schneidman, Berry, Segev, Bialek (2023.06.21.545848v2.36) 2006; 440
Panzeri, Moroni, Safaai, Harvey (2023.06.21.545848v2.43) 2022; 23
Maddison, Mnih, Teh (2023.06.21.545848v2.81) 2016
Dechant, Sasa (2023.06.21.545848v2.76) 2020; 117
Vertes, Sahani (2023.06.21.545848v2.19) 2018; 31
Blanco Malerba, Pieropan, Burak, Azeredo da Silveira (2023.06.21.545848v2.53) 2022
Morais, Pillow (2023.06.21.545848v2.59) 2018; 31
Atick, Redlich (2023.06.21.545848v2.3) 1990; 2
Wei, Stocker (2023.06.21.545848v2.55) 2015; 18
Laughlin, de Ruyter van Steveninck, Anderson (2023.06.21.545848v2.8) 1998; 1
Zhang, Sejnowski (2023.06.21.545848v2.75) 1999; 11
Lee, Mumford (2023.06.21.545848v2.21) 2003; 20
Brunel, Nadal (2023.06.21.545848v2.31) 1998; 10
Aitchison, Hennequin, Lengyel (2023.06.21.545848v2.65) 2018
Seung, Sompolinsky (2023.06.21.545848v2.74) 1993; 90
Wei, Stocker (2023.06.21.545848v2.54) 2012; 25
Shu, Bui, Zhao, Kochenderfer, Ermon (2023.06.21.545848v2.68) 2018; 231
Kingma, Welling (2023.06.21.545848v2.25) 2019; 12
Goodfellow, Bengio, Courville (2023.06.21.545848v2.41) 2016
Salinas, Abbott (2023.06.21.545848v2.44) 1994; 1
Alemi, Poole, Fischer, Dillon, Saurous, Murphy (2023.06.21.545848v2.28) 2018; 35
Dalal, Hall (2023.06.21.545848v2.40) 1983; 45
Harper, McAlpine (2023.06.21.545848v2.32) 2004; 430
Runyan, Piasini, Panzeri, Harvey (2023.06.21.545848v2.38) 2017; 548
Prat-Carrabin, Woodford (2023.06.21.545848v2.12) 2021; 34
Gjorgjieva, Sompolinsky, Meister (2023.06.21.545848v2.34) 2014; 34
Olshausen, Field (2023.06.21.545848v2.64) 1996; 381
Sreenivasan, Fiete (2023.06.21.545848v2.52) 2011; 14
Tomczak, Welling (2023.06.21.545848v2.49) 2017
Lange, Shivkumar, Chattoraj, Haefner (2023.06.21.545848v2.62) 2023; 26
Berkes, Orbán, Lengyel, Fiser (2023.06.21.545848v2.71) 2011; 331
Park, Pillow (2023.06.21.545848v2.11) 2020
References_xml – volume: 7
  start-page: 889
  issue: 5
  year: 1995
  end-page: 904
  ident: 2023.06.21.545848v2.14
  article-title: The Helmholtz Machine
  publication-title: Neural Computation
  doi: 10.1162/neco.1995.7.5.889
– volume: 35
  year: 2018
  ident: 2023.06.21.545848v2.28
  article-title: Fixing a Broken ELBO
  publication-title: International Conference on Machine Learning
– volume: 13
  issue: 1
  year: 2022
  ident: 2023.06.21.545848v2.27
  article-title: Efficient neural codes naturally emerge through gradient descent learning
  publication-title: Nature Communications
  doi: 10.1038/s41467-022-35659-7
– volume: 117
  start-page: 6430
  issue: 12
  year: 2020
  end-page: 6436
  ident: 2023.06.21.545848v2.76
  article-title: Fluctuation-response inequality out of equilibrium
  publication-title: Proceedings of the National Academy of Sciences
  doi: 10.1073/pnas.1918386117/-/DCSupplemental.y
– volume: 231
  start-page: 4393
  year: 2018
  end-page: 4402
  ident: 2023.06.21.545848v2.68
  article-title: Amortized Inference Regularization
  publication-title: Advances in Neural Information Processing Systems
– year: 2016
  ident: 2023.06.21.545848v2.57
  article-title: Neural and perceptual signatures of efficient sensory coding
  publication-title: arXiv preprint
– volume: 11
  start-page: 75
  issue: 1
  year: 1999
  end-page: 84
  ident: 2023.06.21.545848v2.75
  article-title: Neuronal Tuning: To Sharpen or Broaden?
  publication-title: Neural Computation
  doi: 10.1162/089976699300016809
– volume: 12
  start-page: 1
  issue: 1
  year: 2021
  end-page: 14
  ident: 2023.06.21.545848v2.66
  article-title: Unsupervised deep learning identifies semantic disentanglement in single inferotemporal face patch neurons
  publication-title: Nature Communications
  doi: 10.1038/s41467-021-26751-5
– volume: 15
  start-page: 293
  year: 2003
  end-page: 300
  ident: 2023.06.21.545848v2.22
  article-title: Interpreting Neural Response Variability as Monte Carlo Sampling of the Posterior
  publication-title: Advances in Neural Information Processing Systems
– year: 2022
  ident: 2023.06.21.545848v2.60
  article-title: A unifying theory explains seemingly contradicting biases in perceptual estimation
  publication-title: bioRxiv
  doi: 10.1101/2022.12.12.519538
– volume: 92
  start-page: 530
  issue: 2
  year: 2016
  end-page: 543
  ident: 2023.06.21.545848v2.17
  article-title: Neural Variability and Sampling-Based Probabilistic Representations in the Visual Cortex
  publication-title: Neuron
  doi: 10.1016/j.neuron.2016.09.038
– volume: 28
  start-page: 305
  issue: 2
  year: 2016
  end-page: 326
  ident: 2023.06.21.545848v2.58
  article-title: Mutual information, fisher information, and efficient coding
  publication-title: Neural Computation
  doi: 10.1162/NECOa00804
– volume: 36
  start-page: 910
  issue: 9-10
  year: 1981
  end-page: 912
  ident: 2023.06.21.545848v2.2
  article-title: A simple coding procedure enhances a neuron’s information capacity
  publication-title: Zeitschrift für Naturforschung
  doi: 10.1515/znc-1981-9-1040
– start-page: 399246
  year: 2018
  ident: 2023.06.21.545848v2.26
  article-title: Sparse-Coding Variational Auto-Encoders
  publication-title: bioRxiv
  doi: 10.1101/399246
– volume: 14
  start-page: 1330
  issue: 10
  year: 2011
  end-page: 1337
  ident: 2023.06.21.545848v2.52
  article-title: Grid cells generate an analog error-correcting code for singularly precise neural computation
  publication-title: Nature Neuroscience
  doi: 10.1038/nn.2901
– volume: 25
  start-page: 1304
  year: 2012
  end-page: 1312
  ident: 2023.06.21.545848v2.54
  article-title: Efficient coding provides a direct link between prior and likelihood in perceptual Bayesian inference
  publication-title: Advances in Neural Information Processing Systems
– year: 2016
  ident: 2023.06.21.545848v2.82
  article-title: Categorical reparameterization with gumbel-softmax
  publication-title: arXiv preprint
– volume: 2
  year: 2014
  ident: 2023.06.21.545848v2.23
  article-title: Auto-Encoding Variational Bayes
  publication-title: International Conference on Learning Representations
  doi: 10.48550/arXiv.1312.6114
– volume: 130
  start-page: 3880
  year: 2021
  end-page: 3888
  ident: 2023.06.21.545848v2.70
  article-title: Rate-Regularization and Generalization in Variational Autoencoders
  publication-title: Proceedings of The 24th International Conference on Artificial Intelligence and Statistics
– volume: 171
  start-page: 157
  issue: 2
  year: 1992
  end-page: 170
  ident: 2023.06.21.545848v2.4
  article-title: Theoretical predictions of spatiotemporal receptive fields of fly LMCs, and experimental validation
  publication-title: Journal of Comparative Physiology A
  doi: 10.1007/BF00188924
– volume: 331
  start-page: 83
  issue: 6013
  year: 2011
  end-page: 88
  ident: 2023.06.21.545848v2.71
  article-title: Spontaneous cortical activity reveals hallmarks of an optimal internal model of the environment
  publication-title: Science
  doi: 10.1126/science.1195870
– volume: 1
  start-page: 89
  issue: 1
  year: 1994
  end-page: 107
  ident: 2023.06.21.545848v2.44
  article-title: Vector reconstruction from firing rates
  publication-title: Journal of Computational Neuroscience
  doi: 10.1007/BF00962720
– volume: 30
  start-page: 3745
  year: 2016
  end-page: 3753
  ident: 2023.06.21.545848v2.46
  article-title: Ladder Variational Autoencoders
  publication-title: Advances in Neural Information Processing Systems
  doi: 10.48550/arXiv.1602.02282
– volume: 40
  start-page: 2897
  issue: 12
  year: 2018
  end-page: 2905
  ident: 2023.06.21.545848v2.50
  article-title: Information Dropout: Learning Optimal Representations Through Noisy Computation
  publication-title: IEEE Transactions on Pattern Analysis and Machine Intelligence
  doi: 10.1109/TPAMI.2017.2784440
– year: 2004
  ident: 2023.06.21.545848v2.78
  publication-title: Convex optimization
– volume: 16
  start-page: e1008146
  issue: 9
  year: 2020
  ident: 2023.06.21.545848v2.56
  article-title: Efficient sensory coding of multidimensional stimuli
  publication-title: PLoS Computational Biology
  doi: 10.1371/journal.pcbi.1008146
– volume: 18
  start-page: 1509
  issue: 10
  year: 2015
  end-page: 1517
  ident: 2023.06.21.545848v2.55
  article-title: A Bayesian observer model constrained by efficient coding can explain ‘anti-Bayesian’ percepts
  publication-title: Nature Neuroscience
  doi: 10.1038/nn.4105
– volume: 107
  start-page: 14419
  issue: 32
  year: 2010
  end-page: 14424
  ident: 2023.06.21.545848v2.37
  article-title: Optimal population coding by noisy spiking neurons
  publication-title: PNAS
  doi: 10.1073/pnas.1004906107/-/DCSupplemental
– year: 2022
  ident: 2023.06.21.545848v2.18
  article-title: Top-down inference in an early visual cortex inspired hierarchical Variational Autoencoder
  publication-title: arXiv preprint
– volume: 54
  start-page: 610
  issue: 3
  year: 1973
  end-page: 619
  ident: 2023.06.21.545848v2.61
  article-title: Frequency difference limens for short-duration tones
  publication-title: The Journal of the Acoustical Society of America
– volume: 360
  start-page: 68
  issue: 6399
  year: 1992
  end-page: 70
  ident: 2023.06.21.545848v2.5
  article-title: Real and optimal neural images in early vision
  publication-title: Nature
  doi: 10.1038/360068a0
– volume: 7
  start-page: 1135
  issue: 7
  year: 2023
  end-page: 1151
  ident: 2023.06.21.545848v2.30
  article-title: Sensory perception relies on fitness-maximizing codes
  publication-title: Nature Human Behaviour
  doi: 10.1038/s41562-023-01584-y
– volume: 10
  start-page: 1731
  issue: 7
  year: 1998
  end-page: 1757
  ident: 2023.06.21.545848v2.31
  article-title: Mutual Information, Fisher Information, and Population Coding
  publication-title: Neural Computation
  doi: 10.1162/089976698300017115
– start-page: 2020
  year: 2020
  ident: 2023.06.21.545848v2.67
  article-title: Adaptive Efficient Coding: A Variational Auto-encoder Approach
  publication-title: bioRxiv
  doi: 10.1101/2020.05.29.124453
– volume: 34
  start-page: 12127
  issue: 36
  year: 2014
  end-page: 12144
  ident: 2023.06.21.545848v2.34
  article-title: Benefits of pathway splitting in sensory coding
  publication-title: Journal of Neuroscience
  doi: 10.1523/JNEUROSCI.1032-14.2014
– volume: 10
  start-page: 403
  issue: 2
  year: 1998
  end-page: 430
  ident: 2023.06.21.545848v2.20
  article-title: Probabilistic Interpretation of Population Codes
  publication-title: Neural Computation
  doi: 10.1162/089976698300017818
– volume: 31
  start-page: 1278
  year: 2014
  end-page: 1286
  ident: 2023.06.21.545848v2.24
  article-title: Stochastic backpropagation and approximate inference in deep generative models
  publication-title: International conference on machine learning
  doi: 10.48550/arXiv.1401.4082
– start-page: 2022
  year: 2022
  ident: 2023.06.21.545848v2.53
  article-title: Random Compressed Coding with Neurons
  publication-title: bioRxiv
  doi: 10.1101/2022.01.06.475186
– year: 2018
  ident: 2023.06.21.545848v2.65
  article-title: Sampling-based probabilistic inference emerges from learning in neural circuits with a cost on reliability
  publication-title: arXiv preprint
– volume: 34
  start-page: 23793
  year: 2021
  end-page: 23805
  ident: 2023.06.21.545848v2.12
  article-title: Bias and variance of the Bayesian-mean decoder
  publication-title: Advances in Neural Information Processing Systems
– volume: 548
  start-page: 92
  issue: 7665
  year: 2017
  end-page: 96
  ident: 2023.06.21.545848v2.38
  article-title: Distinct timescales of population coding across cortex
  publication-title: Nature
  doi: 10.1038/nature23020
– year: 2016
  ident: 2023.06.21.545848v2.83
  article-title: Discrete Variational Autoencoders
  publication-title: arXiv preprint
  doi: 10.48550/arXiv.1609.02200
– volume: 29
  start-page: 37
  year: 2019
  end-page: 44
  ident: 2023.06.21.545848v2.45
  article-title: Linking neural responses to behavior with information-preserving population vectors
  publication-title: Current Opinion in Behavioral Sciences
  doi: 10.1016/j.cobeha.2019.03.004
– volume: 31
  start-page: 5071
  year: 2018
  end-page: 5080
  ident: 2023.06.21.545848v2.59
  article-title: Power-law efficient neural codes provide general link between perceptual bias and discriminability
  publication-title: Advances in Neural Information Processing Systems
– year: 2014
  ident: 2023.06.21.545848v2.7
  publication-title: Theory, Models, and Data
– volume: 15
  start-page: 1
  issue: 11
  year: 2019
  end-page: 38
  ident: 2023.06.21.545848v2.35
  article-title: Functional diversity among sensory neurons from efficient coding principles
  publication-title: PLoS Computational Biology
  doi: 10.1371/journal.pcbi.1007476
– volume: 3
  year: 2015
  ident: 2023.06.21.545848v2.48
  article-title: Adam: A Method for Stochastic Optimization
  publication-title: International Conference on Learning Representations
  doi: 10.48550/arXiv.1412.6980
– year: 2016
  ident: 2023.06.21.545848v2.81
  article-title: The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
  publication-title: arXiv preprint
– volume: 25
  year: 2012
  ident: 2023.06.21.545848v2.29
  article-title: Optimal Neural Tuning Curves for Arbitrary Stimulus Distributions: Discrimax, Infomax and Minimum L p Loss
  publication-title: Advances in neural information processing systems
– volume: 9
  start-page: 1432
  issue: 11
  year: 2006
  end-page: 1438
  ident: 2023.06.21.545848v2.73
  article-title: Bayesian inference with probabilistic population codes
  publication-title: Nature Neuroscience
  doi: 10.1038/nn1790
– volume: 4
  issue: 11
  year: 1927
  ident: 2023.06.21.545848v2.13
  article-title: Helmholtz’s treatise on physiological optics
  publication-title: Optometry and Vision Science
  doi: 10.1097/00006324-192711000-00011
– volume: 23
  start-page: 551
  issue: 9
  year: 2022
  end-page: 567
  ident: 2023.06.21.545848v2.43
  article-title: The structures and functions of correlations in neural population codes
  publication-title: Nature Reviews Neuroscience
– volume: 31
  start-page: 4166
  year: 2018
  end-page: 4175
  ident: 2023.06.21.545848v2.19
  article-title: Flexible and accurate inference and learning for deep generative models
  publication-title: Advances in Neural Information Processing Systems
– volume: 5
  year: 2017
  ident: 2023.06.21.545848v2.47
  article-title: beta-VAE: Learning basic visual concepts with a constrained variational framework
  publication-title: International Conference on Learning Representations
– start-page: 1214
  year: 2017
  end-page: 1223
  ident: 2023.06.21.545848v2.49
  article-title: VAE with a VampPrior
  publication-title: International Conference on Artificial Intelligence and Statistics
– volume: 44
  start-page: 403
  year: 2021
  end-page: 424
  ident: 2023.06.21.545848v2.42
  article-title: The Geometry of Information Coding in Correlated Neural Populations
  publication-title: Annual Review of Neuroscience
  doi: 10.1146/annurev-neuro-120320-082744
– volume: 440
  start-page: 1007
  issue: 7087
  year: 2006
  end-page: 1012
  ident: 2023.06.21.545848v2.36
  article-title: Weak pairwise correlations imply strongly correlated network states in a neural population
  publication-title: Nature
  doi: 10.1038/nature04701
– volume: 20
  start-page: 2526
  issue: 10
  year: 2008
  end-page: 2563
  ident: 2023.06.21.545848v2.63
  article-title: Sparse coding via thresholding and local competition in neural circuits
  publication-title: Neural Computation
  doi: 10.1162/neco.2008.03-07-486
– volume: 90
  start-page: 10749
  issue: 22
  year: 1993
  end-page: 10753
  ident: 2023.06.21.545848v2.74
  article-title: Simple models for reading neuronal population codes
  publication-title: Proceedings of the National Academy of Sciences of the United States of America
  doi: 10.1073/pnas.90.22.10749
– volume: 1
  start-page: 216
  issue: 1
  year: 1961
  end-page: 234
  ident: 2023.06.21.545848v2.1
  article-title: Possible Principles Underlying the Transformations of Sensory Messages
  publication-title: Sensory Communication
  doi: 10.7551/mitpress/9780262518420.003.0013
– volume: 1
  start-page: 36
  issue: 1
  year: 1998
  end-page: 41
  ident: 2023.06.21.545848v2.8
  article-title: The metabolic cost of neural information
  publication-title: Nature Neuroscience
  doi: 10.1038/236
– volume: 14
  start-page: 2317
  issue: 10
  year: 2002
  end-page: 2351
  ident: 2023.06.21.545848v2.51
  article-title: Optimal Short-Term Population Coding: When Fisher Information Fails
  publication-title: Neural Computation
  doi: 10.1162/08997660260293247
– volume: 430
  start-page: 682
  issue: 7000
  year: 2004
  end-page: 686
  ident: 2023.06.21.545848v2.32
  article-title: Optimal neural population coding of an auditory spatial cue
  publication-title: Nature
  doi: 10.1038/nature02768
– volume: 26
  start-page: 2063
  issue: 12
  year: 2023
  end-page: 2072
  ident: 2023.06.21.545848v2.62
  article-title: Bayesian encoding and decoding as distinct perspectives on neural coding
  publication-title: Nature Neuroscience
  doi: 10.1038/s41593-023-01458-6
– volume: 2
  year: 1958
  ident: 2023.06.21.545848v2.80
  publication-title: Studies in linear and non-linear programming
– volume: 2
  start-page: 308
  issue: 3
  year: 1990
  end-page: 320
  ident: 2023.06.21.545848v2.3
  article-title: Towards a Theory of Early Visual Processing
  publication-title: Neural Computation
  doi: 10.1162/neco.1990.2.3.308
– volume: 32
  year: 2018
  ident: 2023.06.21.545848v2.69
  article-title: Bias and Generalization in Deep Generative Models: An Empirical Study
  publication-title: Advances in Neural Information Processing Systems
– volume: 37
  year: 2020
  ident: 2023.06.21.545848v2.79
  article-title: On gradient descent ascent for nonconvex-concave minimax problems
  publication-title: International Conference on Machine Learning
  doi: 10.48550/arXiv.1906.00331
– volume: 26
  start-page: 2103
  issue: 10
  year: 2014
  end-page: 2134
  ident: 2023.06.21.545848v2.10
  article-title: Efficient Sensory Encoding and Bayesian Inference with Heterogeneous Neural Populations
  publication-title: Neural Computation
  doi: 10.1162/NECOa00638
– volume: 142
  start-page: 205
  issue: 1
  year: 2009
  end-page: 228
  ident: 2023.06.21.545848v2.77
  article-title: Subgradient methods for saddle-point problems
  publication-title: Journal of Optimization Theory and Applications
  doi: 10.1007/s10957-009-9522-7
– volume: 45
  issue: 2
  year: 1983
  ident: 2023.06.21.545848v2.40
  article-title: Approximating Priors by Mixtures of Natural Conjugate Priors
  publication-title: Journal of the Royal Statistical Society: Series B (Methodological)
  doi: 10.1111/j.2517-6161.1983.tb01251.x
– start-page: 178418
  year: 2020
  ident: 2023.06.21.545848v2.11
  article-title: Bayesian Efficient Coding
  publication-title: bioRxiv
  doi: 10.1101/178418
– volume: 7552
  start-page: 523
  year: 2012
  end-page: 530
  ident: 2023.06.21.545848v2.33
  article-title: Bayesian inference with efficient neural population codes
  publication-title: In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
– volume: 5
  start-page: 356
  issue: 4
  year: 2002
  end-page: 363
  ident: 2023.06.21.545848v2.6
  article-title: Efficient coding of natural sounds
  publication-title: Nature Neuroscience
  doi: 10.1038/nn831
– volume: 20
  start-page: 1434
  issue: 7
  year: 2003
  end-page: 1448
  ident: 2023.06.21.545848v2.21
  article-title: Hierarchical Bayesian inference in the visual cortex
  publication-title: JOSA A
  doi: 10.1364/JOSAA.20.001434
– volume: 23
  start-page: 122
  issue: 1
  year: 2020
  end-page: 129
  ident: 2023.06.21.545848v2.39
  article-title: A neural basis of probabilistic computation in visual cortex
  publication-title: Nature Neuroscience
  doi: 10.1038/s41593-019-0554-5
– year: 2001
  ident: 2023.06.21.545848v2.15
  publication-title: Computational and Mathematical Modeling of Neural Systems
– year: 2016
  ident: 2023.06.21.545848v2.41
  publication-title: Deep learning
– volume: 21
  start-page: 1133
  issue: 10
  year: 2001
  end-page: 1145
  ident: 2023.06.21.545848v2.9
  article-title: An energy budget for signaling in the grey matter of the brain
  publication-title: Journal of Cerebral Blood Flow and Metabolism
  doi: 10.1097/00004647-200110000-00001
– volume: 381
  start-page: 607
  issue: 6583
  year: 1996
  end-page: 609
  ident: 2023.06.21.545848v2.64
  article-title: Emergence of simple-cell receptive field properties by learning a sparse code for natural images
  publication-title: Nature
  doi: 10.1038/381607a0
– volume: 90
  start-page: 649
  issue: 3
  year: 2016
  end-page: 660
  ident: 2023.06.21.545848v2.72
  article-title: Perceptual Decision-Making as Probabilistic Inference by Neural Sampling
  publication-title: Neuron
  doi: 10.1016/j.neuron.2016.03.020
– volume: 12
  start-page: 307
  issue: 4
  year: 2019
  end-page: 392
  ident: 2023.06.21.545848v2.25
  article-title: An introduction to variational autoencoders
  publication-title: Foundations and Trends in Machine Learning
  doi: 10.1561/2200000056
– volume: 12
  start-page: 855
  year: 1999
  end-page: 861
  ident: 2023.06.21.545848v2.16
  article-title: Scale mixtures of Gaussians and the statistics of natural images
  publication-title: Advances in Neural Information Processing Systems
SSID ssj0002961374
Score 1.721186
SecondaryResourceType preprint
Snippet The efficient coding approach proposes that neural systems represent as much sensory information as biological constraints allow. It aims at formalizing...
SourceID biorxiv
SourceType Open Access Repository
SubjectTerms Neuroscience
Title Jointly efficient encoding and decoding in neural populations
URI https://www.biorxiv.org/content/10.1101/2023.06.21.545848
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwtV1La9wwEBZLtoXe2rSlTR-40Nvidi3bsnRMQkJow7KkacnNyJYWDMZanI3Z5Nfnk_xY0-SQHnoRtiTLSN8gzYzmQchXmVEwoSvuKxEyP8IR4oOUIx9nI1_FcZLPXY6lP-fJYsGvrsRyMvnZ-8I0ZVJVfLsV6_8KNeoAtnWd_Qe4h0FRgWeAjhKwo3wS8D9MUW3KW2upUThvx5mNVal6Z0Slu5eimtlgltYTa0jidT3mVbPCXGyLZqezLgGB4zV_AWAwp0clSMYM7c6o1BkHHN7Upt7t98ao3oR-bKVvzWclxiobXdQtDysbOTu8s-lDzVgdQSNn_LdTRx6bEnyy00laNyTQ8ey8pWbTmTVrt7lRJkAV1LlcP7KNu_QBNrO7Da5Kg2_2eq-NyPlXdOwHfRqcx1OaxAJb3PToZLG8GFRuVIB3SaLubhv_-P7ga0hBWNwaizviMi5fkulSrnX9ikx0tU-et2lCb1-THlJvgNTrIfUAqddD6hWV10LqjSB9Q36fnlwen_ld2gs_sz4ZPo-zaK5prliy0jTOXQAjoUIrikvJMqZlLLKAowuPBdPMNQchpheGjOvwLdmrQAXviAdpWSdJQOdZpCLJId3muaLophhTUoTvyZduuum6DW6S2iVJ5yylQdouycET-nwgL3a08JHsbeob_Yk8y5tNcV1_7nC4ByN8QEc
linkProvider ProQuest
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Jointly+efficient+encoding+and+decoding+in+neural+populations&rft.jtitle=bioRxiv&rft.au=Malerba%2C+Simone+Blanco&rft.au=Micheli%2C+Aurora&rft.au=Woodford%2C+Michael&rft.au=da+Silveira%2C+Rava+Azeredo&rft.date=2024-05-04&rft.pub=Cold+Spring+Harbor+Laboratory&rft.eissn=2692-8205&rft_id=info:doi/10.1101%2F2023.06.21.545848&rft.externalDocID=2023.06.21.545848v2