CoChat

Hybrid predictive coding: Inferring, fast and slow

Open access
Tscshantz, Alexander, Millidge, Beren, Seth, Anil K., Buckley, Christopher L. · 2023 · PLOS Computational Biology
Cited by 51

The DOI matches its Crossref record; 54 of the 54 references it cites were found in a registry or an index.

Abstract

Predictive coding is an influential model of cortical neural activity. It proposes that perceptual beliefs are furnished by sequentially minimising “prediction errors”—the differences between predicted and observed data. Implicit in this proposal is the idea that successful perception requires multiple cycles of neural activity. This is at odds with evidence that several aspects of visual perception—including complex forms of object recognition—arise from an initial “feedforward sweep” that occurs on fast timescales which preclude substantial recurrent activity. Here, we propose that the feedforward sweep can be understood as performing amortized inference (applying a learned function that maps directly from data to beliefs) and recurrent processing can be understood as performing iterative inference (sequentially updating neural activity in order to improve the accuracy of beliefs). We propose a hybrid predictive coding network that combines both iterative and amortized inference in a principled manner by describing both in terms of a dual optimization of a single objective function. We show that the resulting scheme can be implemented in a biologically plausible neural architecture that approximates Bayesian inference utilising local Hebbian update rules. We demonstrate that our hybrid predictive coding model combines the benefits of both amortized and iterative inference—obtaining rapid and computationally cheap perceptual inference for familiar data while maintaining the context-sensitivity, precision, and sample efficiency of iterative inference schemes. Moreover, we show how our model is inherently sensitive to its uncertainty and adaptively balances iterative and amortized inference to obtain accurate beliefs using minimum computational expense. Hybrid predictive coding offers a new perspective on the functional relevance of the feedforward and recurrent activity observed during visual perception and offers novel insights into distinct aspects of visual phenomenology.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1371/journal.pcbi.1011280

Cites

54 of 54 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 54 references

Research works printed but not linked: 91

  1. P Dayan, 1995, The Helmholtz machine, Neural computation, 7, 889, 10.1162/neco.1995.7.5.889 · DOI · Paper page
  2. K Friston, 2005, A theory of cortical responses, Philosophical transactions of the Royal Society B: Biological sciences, 360, 815, 10.1098/rstb.2005.1622 · DOI · Paper page
  3. R Weidner, 2009, Sources of top–down control in visual search, Journal of Cognitive Neuroscience, 21, 2100, 10.1162/jocn.2008.21173 · DOI · Paper page
  4. L Melloni, 2012, Interaction between bottom-up saliency and top-down control: how saliency maps are created in the human brain, Cerebral cortex, 22, 2943, 10.1093/cercor/bhr384 · DOI
  5. A Delorme, 2004, Interaction of top-down and bottom-up processing in the fast visual analysis of natural scenes, Cognitive Brain Research, 19, 103, 10.1016/j.cogbrainres.2003.11.010 · DOI · Paper page
  6. G Kreiman, 2020, Beyond the feedforward sweep: feedback computations in the visual cortex, Annals of the New York Academy of Sciences
  7. M Ahissar, 2004, The reverse hierarchy theory of visual perceptual learning, Trends in cognitive sciences, 8, 457, 10.1016/j.tics.2004.08.011 · DOI · Paper page
  8. R VanRullen, 2007, The power of the feed-forward sweep, Advances in Cognitive Psychology, 3, 167, 10.2478/v10053-008-0022-3 · DOI · Paper page
  9. PE Roland, 2010, Six principles of visual cortical dynamics, Frontiers in systems neuroscience, 4, 28, 10.3389/fnsys.2010.00028 · DOI
  10. K Rauss, 2013, What is bottom-up and what is top-down in predictive coding?, Frontiers in psychology, 4, 276, 10.3389/fpsyg.2013.00276 · DOI · Paper page
  11. N Gordon, 2017, Neural markers of predictive coding under perceptual uncertainty revealed with Hierarchical Frequency Tagging, Elife, 6, e22749, 10.7554/eLife.22749 · DOI · Paper page
  12. CL Buckley, 2017, The free energy principle for action and perception: A mathematical review, Journal of Mathematical Psychology, 81, 55, 10.1016/j.jmp.2017.09.004 · DOI · Paper page
  13. Millidge B, Seth AK, Buckley CL. Predictive coding: a theoretical and experimental review. arXiv preprint arXiv:210712979. 2021;.
  14. A Clark, 2015, Surfing uncertainty: Prediction, action, and the embodied mind
  15. Millidge B. Implementing predictive processing and active inference: Preliminary steps and results. PsyArXiv. 2019;. · DOI
  16. CW Fox, 2012, A tutorial on variational Bayesian inference, Artificial intelligence review, 38, 85, 10.1007/s10462-011-9236-8 · DOI · Paper page
  17. Millidge B, Tschantz A, Seth AK, Buckley C. Neural Kalman filtering. arXiv preprint arXiv:210210021. 2021;.
  18. Millidge B, Tschantz A, Seth AK, Buckley CL. Reinforcement Learning as Iterative and Amortised Inference. arXiv preprint arXiv:200610524. 2020;.
  19. Tschantz A, Millidge B, Seth AK, Buckley CL. Control as hybrid inference. arXiv preprint arXiv:200705838. 2020;.
  20. Marino J, Yue Y, Mandt S. Iterative amortized inference. In: International Conference on Machine Learning. PMLR; 2018. p. 3403–3412.
  21. van Bergen RS, Kriegeskorte N. Going in circles is the way forward: the role of recurrence in visual inference. arXiv preprint arXiv:200312128. 2020;. · DOI · Paper page
  22. C Keysers, 2001, The speed of sight, Journal of cognitive neuroscience, 13, 90, 10.1162/089892901564199 · DOI · Paper page
  23. T Carlson, 2013, Representational dynamics of object vision: the first 1000 ms, Journal of vision, 13, 1, 10.1167/13.10.1 · DOI · Paper page
  24. E Thunell, 2019, Memory for repeated images in rapid-serial-visual-presentation streams of thousands of images, Psychological science, 30, 989, 10.1177/0956797619842251 · DOI
  25. Kingma DP, Welling M. Auto-encoding variational bayes. arXiv preprint arXiv:13126114. 2013;.
  26. Doersch C. Tutorial on variational autoencoders. arXiv preprint arXiv:160605908. 2016;.
  27. Gershman S, Goodman N. Amortized inference in probabilistic reasoning. In: Proceedings of the annual meeting of the cognitive science society. vol. 36; 2014.
  28. Cremer C, Li X, Duvenaud D. Inference suboptimality in variational autoencoders. In: International Conference on Machine Learning. PMLR; 2018. p. 1078–1086.
  29. D Kahneman, 2011, Thinking, fast and slow
  30. Ghahramani Z, Beal MJ. Propagation algorithms for variational Bayesian learning. In: Advances in neural information processing systems; 2001. p. 507–513.
  31. Marino J, Piché A, Ialongo AD, Yue Y. Iterative amortized policy optimization. Advances in Neural Information Processing Systems. 2021;34.
  32. Millidge B, Tschantz A, Seth AK, Buckley CL. On the relationship between active inference and control as inference. In: International Workshop on Active Inference. Springer; 2020. p. 3–11. · DOI
  33. Schmidhuber J. Making the World Differentiable: On Using Self-Supervised Fully Recurrent Neural Networks for Dynamic Reinforcement Learning and Planning in Non-Stationary Environments. Technical Report: Technical University of Munich. 1990;. · DOI · Paper page
  34. Bengio Y, Scellier B, Bilaniuk O, Sacramento J, Senn W. Feedforward initialization for fast inference of deep generative networks is biologically plausible. arXiv preprint arXiv:160601651. 2016;.
  35. X Xie, 2003, Equivalence of backpropagation and contrastive Hebbian learning in a layered network, Neural computation, 15, 441, 10.1162/089976603762552988 · DOI · Paper page
  36. B Scellier, 2017, Equilibrium propagation: Bridging the gap between energy-based models and backpropagation, Frontiers in computational neuroscience, 11, 24, 10.3389/fncom.2017.00024 · DOI · Paper page
  37. RP Rao, 1999, Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects, Nature Neuroscience, 2, 79, 10.1038/4580 · DOI · Paper page
  38. D Xu, 2017, Symmetric predictive estimator for biologically plausible neural learning, IEEE Transactions on Neural Networks and Learning Systems, 29, 4140, 10.1109/TNNLS.2017.2756859 · DOI
  39. K Friston, 2012, The history of the future of the Bayesian brain, NeuroImage, 62, 1230, 10.1016/j.neuroimage.2011.10.004 · DOI · Paper page
  40. DC Knill, 1996, Perception as Bayesian inference, 10.1017/CBO9780511984037 · DOI · Paper page
  41. RT Cox, 1946, Probability, frequency and reasonable expectation, American journal of physics, 14, 1, 10.1119/1.1990764 · DOI · Paper page
  42. MJ Wainwright, 2008, Graphical models, exponential families, and variational inference, Foundations and Trends in Machine Learning, 1, 1, 10.1561/2200000001 · DOI · Paper page
  43. Odaibo S. Tutorial: Deriving the standard variational autoencoder (vae) loss function. arXiv preprint arXiv:190708956. 2019;.
  44. MD Hoffman, 2013, Stochastic variational inference, The Journal of Machine Learning Research, 14, 1303
  45. Ranganath R, Gerrish S, Blei D. Black box variational inference. In: Artificial intelligence and statistics. PMLR; 2014. p. 814–822.
  46. Millidge B, Tschantz A, Seth AK, Buckley CL. Relaxing the constraints on predictive coding models. arXiv preprint arXiv:201001047. 2020;.
  47. C Zhang, 2018, Advances in variational inference, IEEE transactions on pattern analysis and machine intelligence, 41, 2008, 10.1109/TPAMI.2018.2889774 · DOI · Paper page
  48. I Goodfellow, 2016, Deep learning
  49. Kingma DP, Ba J. Adam: A method for stochastic optimization. arXiv preprint arXiv:14126980. 2014;.
  50. Kinghorn PF, Millidge B, Buckley CL. Preventing Deterioration of Classification Accuracy in Predictive Coding Networks. arXiv preprint arXiv:220807114. 2022;. · DOI
  51. K Doya, 2007, Bayesian brain: Probabilistic approaches to neural coding
  52. Wolpert DM, Ghahramani Z. Bayes rule in perception, action and cognition. The Oxford Companion to the Mind Oxford University Press (http://eprints.pascal-network.org/archive/00001354/). 2005;.
  53. MJ Ramstead, 2022, From generative models to generative passages: a computational approach to (neuro) phenomenology, Review of Philosophy and Psychology, 1
  54. K Suzuki, 2017, A deep-dream virtual reality platform for studying altered perceptual phenomenology, Scientific reports, 7, 1, 10.1038/s41598-017-16316-2 · DOI · Paper page
  55. MW Spratling, 2012, Unsupervised learning of generative and discriminative weights encoding elementary image components in a predictive coding model of cortical function, Neural computation, 24, 60, 10.1162/NECO_a_00222 · DOI · Paper page
  56. JS Johnson, 2005, The recognition of partially visible natural objects in the presence and absence of their occluders, Vision research, 45, 3262, 10.1016/j.visres.2005.06.007 · DOI
  57. D Wyatte, 2012, The limits of feedforward vision: Recurrent processing promotes robust object recognition when objects are degraded, Journal of Cognitive Neuroscience, 24, 2248, 10.1162/jocn_a_00282 · DOI · Paper page
  58. K Rajaei, 2019, Beyond core object recognition: Recurrent processes account for object recognition under occlusion, PLoS computational biology, 15, e1007001, 10.1371/journal.pcbi.1007001 · DOI · Paper page
  59. H Tang, 2018, Recurrent computations for visual pattern completion, Proceedings of the National Academy of Sciences, 115, 8835, 10.1073/pnas.1719397115 · DOI · Paper page
  60. S Grossberg, 2013, Adaptive Resonance Theory: How a brain learns to consciously attend, learn, and recognize a changing world, Neural networks, 37, 1, 10.1016/j.neunet.2012.09.017 · DOI · Paper page
  61. T Serre, 2007, A feedforward architecture accounts for rapid categorization, Proceedings of the national academy of sciences, 104, 6424, 10.1073/pnas.0700622104 · DOI · Paper page
  62. SL Brincat, 2006, Dynamic shape synthesis in posterior inferotemporal cortex, Neuron, 49, 17, 10.1016/j.neuron.2005.11.026 · DOI · Paper page
  63. DL Ringach, 1996, Spatial and temporal properties of illusory contours and amodal boundary completion, Vision research, 36, 3037, 10.1016/0042-6989(96)00062-4 · DOI · Paper page
  64. Y Sugase, 1999, Global and fine information coded by single neurons in the temporal visual cortex, Nature, 400, 869, 10.1038/23703 · DOI · Paper page
  65. Kim Y, Wiseman S, Miller A, Sontag D, Rush A. Semi-amortized variational autoencoders. In: International Conference on Machine Learning. PMLR; 2018. p. 2678–2687.
  66. Huang Y, Gornet J, Dai S, Yu Z, Nguyen T, Tsao D, et al. Neural Networks with Recurrent Generative Feedback. Advances in Neural Information Processing Systems. 2020;33.
  67. J Gordon, 2020, Combining deep generative and discriminative models for Bayesian semi-supervised learning, Pattern Recognition, 100, 107156, 10.1016/j.patcog.2019.107156 · DOI
  68. T Grootswagers, 2019, The representational dynamics of visual objects in rapid serial visual processing streams, NeuroImage, 188, 668, 10.1016/j.neuroimage.2018.12.046 · DOI
  69. SA Yoo, 2019, Feed-forward visual processing suffices for coarse localization but fine-grained localization in an attention-demanding context needs feedback processing, Plos one, 14, e0223166, 10.1371/journal.pone.0223166 · DOI
  70. Y Mohsenzadeh, 2018, Ultra-Rapid serial visual presentation reveals dynamics of feedforward and feedback processes in the ventral visual pathway, Elife, 7, e36329, 10.7554/eLife.36329 · DOI
  71. Furtak M, Mudrik L, Bola M. The forest, the trees, or both? Hierarchy and interactions between gist and object processing during perception of real-world scenes. PsyArXiv. 2021;. · DOI
  72. H Karimi-Rouzbahani, 2020, Perceptual difficulty modulates the direction of information flow in familiar face recognition, bioRxiv
  73. CJ Spoerer, 2020, Recurrent neural networks can explain flexible trading of speed and accuracy in biological vision, PLoS computational biology, 16, e1008215, 10.1371/journal.pcbi.1008215 · DOI · Paper page
  74. LA Riggs, 1953, The disappearance of steadily fixated visual test objects, JOSA, 43, 495, 10.1364/JOSA.43.000495 · DOI · Paper page
  75. Millidge B. Fixational Eye Movements: Data Augmentation for the Brain? PsyArXiv. 2019;. · DOI
  76. VA Lamme, 2000, The role of primary visual cortex (V1) in visual awareness, Vision research, 40, 1507, 10.1016/S0042-6989(99)00243-6 · DOI · Paper page
  77. AK Seth, 2021, Being you: A new science of consciousness
  78. J Hohwy, 2008, Predictive coding explains binocular rivalry: An epistemological review, Cognition, 108, 687, 10.1016/j.cognition.2008.05.010 · DOI · Paper page
  79. Lotter W, Kreiman G, Cox D. Deep predictive coding networks for video prediction and unsupervised learning. arXiv preprint arXiv:160508104. 2016;.
  80. Song Y, Lukasiewicz T, Xu Z, Bogacz R. Can the Brain Do Backpropagation?—Exact Implementation of Backpropagation in Predictive Coding Networks. Advances in Neural Information Processing Systems. 2020;33.
  81. A Oliva, 2005, Neurobiology of attention, 251, 10.1016/B978-012375731-9/50045-8 · DOI · Paper page
  82. A Oliva, 2006, Building the gist of a scene: The role of global image features in recognition, Progress in brain research, 155, 23, 10.1016/S0079-6123(06)55002-2 · DOI · Paper page
  83. Kingma DP, Rezende DJ, Mohamed S, Welling M. Semi-supervised learning with deep generative models. arXiv preprint arXiv:14065298. 2014;.
  84. Rezende DJ, Mohamed S, Wierstra D. Stochastic backpropagation and approximate inference in deep generative models. In: International conference on machine learning. PMLR; 2014. p. 1278–1286.
  85. Kuleshov V, Ermon S. Deep hybrid models: Bridging discriminative and generative approaches. In: Proceedings of the Conference on Uncertainty in AI (UAI); 2017.
  86. Liu H, Abbeel P. Hybrid discriminative-generative training via contrastive learning. arXiv preprint arXiv:200709070. 2020;.
  87. V Garcia Satorras, 2019, Combining Generative and Discriminative Models for Hybrid Inference, Advances in Neural Information Processing Systems, 32, 13825
  88. Van de Ven GM, Tolias AS. Generative replay with feedback connections as a general strategy for continual learning. arXiv preprint arXiv:180910635. 2018;.
  89. GM van de Ven, 2020, Brain-inspired replay for continual learning with artificial neural networks, Nature communications, 11, 1, 10.1038/s41467-020-17866-2 · DOI · Paper page
  90. Schmidhuber J. Making the World Differentiable: On Using Self-Supervised Fully Recurrent Neural Networks for Dynamic Reinforcement Learning and Planning in Non-Stationary Environments; 1990. · DOI · Paper page
  91. K Friston, 2017, Active inference, curiosity and insight, Neural computation, 29, 2633, 10.1162/neco_a_00999 · DOI · Paper page

Cited by

See all 51

Metadata

Authors
Tscshantz, A., Millidge, B., Seth, A. K., & Buckley, C. L.
Venue
PLOS Computational Biology
Published
2023-08-02
Volume
19
Issue
8
Pages
e1011280
ISSN
  • 1553-734X (ISSN-L)
  • 1553-7358
OpenAlex
W4385481140
Cited by
51
References
54
FWCI
7.12
Type
journal article
Language
en
Open access
gold
Funders
  • European Research Council (101012954)
  • Canadian Institute for Advanced Research (Program in Brain, Mind, and Consciousness)
  • Dr. Mortimer and Theresa Sackler Foundation
  • School of Engineering and Informatics at the University of Sussex
  • Biotechnology and Biological Sciences Research Council (BB/P022197/1, BB/s006338/1)

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Hybrid predictive coding: Inferring, fast and slow. https://app.cochat.ai/papers/10.1371/journal.pcbi.1011280
The paper
Tscshantz, A., Millidge, B., Seth, A. K., & Buckley, C. L. (2023). Hybrid predictive coding: Inferring, fast and slow. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.1011280

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.