The graphical brain: Belief propagation and active inference
Open accessThe DOI matches its Crossref record; 112 of the 112 references it cites were found in a registry or an index; none of the 73 sources its claims cite is checked for support yet.
Abstract
This paper considers functional integration in the brain from a computational perspective. We ask what sort of neuronal message passing is mandated by active inference—and what implications this has for context-sensitive connectivity at microscopic and macroscopic levels. In particular, we formulate neuronal processing as belief propagation under deep generative models. Crucially, these models can entertain both discrete and continuous states, leading to distinct schemes for belief updating that play out on the same (neuronal) architecture. Technically, we use Forney (normal) factor graphs to elucidate the requisite message passing in terms of its form and scheduling. To accommodate mixed generative models (of discrete and continuous states), one also has to consider link nodes or factors that enable discrete and continuous representations to talk to each other. When mapping the implicit computational architecture onto neuronal connectivity, several interesting features emerge. For example, Bayesian model averaging and comparison, which link discrete and continuous states, may be implemented in thalamocortical loops. These and other considerations speak to a computational connectome that is inherently state dependent and self-organizing in ways that yield to a principled (variational) account. We conclude with simulations of reading that illustrate the implicit neuronal message passing, with a special focus on how discrete (semantic) representations inform, and are informed by, continuous (visual) sampling of the sensorium.
Overview
CoChat Summary
TL;DR
This paper derives belief propagation under deep generative models as a computational architecture for functional integration in the brain, using Forney factor graphs to link discrete and continuous state representations. It maps the resulting message passing onto intrinsic/extrinsic cortical microcircuits and cortico–basal ganglia–thalamic loops, illustrating the scheme with simulations of pictographic reading.
Key findings
- Belief propagation under deep generative models can be expressed with Forney factor graphs; discrete models use A/B/C/D matrices and policies, continuous models use generalized coordinates and predictive coding, and mixed models require link factors implementing Bayesian model averaging and comparison.
- Discrete and continuous representations are linked by Bayesian model averaging (descending empirical priors) and Bayesian model comparison (ascending posterior over outcome models); thalamocortical loops may implement this, while expected free energy and policy expectations involve cortico–basal ganglia–thalamic loops.
- In simulations of reading, the model has two discrete hierarchical levels with six sentences and four words, with letters/pictograms in quadrants; saccades are assumed to take about 256 ms, continuous integration uses 24/25 time steps of about 10 ms, and the subject resolves the sentence on the last saccade of the last word.
- Minimizing variational free energy also minimizes Lyapunov exponents, so belief propagation organizes toward critical slowing; self-organized criticality is described as an epiphenomenon of self-evidencing.
Methods
- Theoretical/computational derivation using Bayesian networks, Forney factor graphs, and neural network/circuit mappings; derives belief propagation for discrete, continuous, and mixed generative models.
- Simulation of pictographic reading: two discrete hierarchical levels (six sentences, four words; letters/pictograms in quadrants), a continuous visuomotor model with 32 × 32 pixel images, equations integrated with 16 iterations per time point per discrete level, saccades about 256 ms, continuous integration 24/25 time steps of about 10 ms, and roughly 4.5 s of simulated time.
- Main analysis is simulation-based: eye movements, posterior probabilities/evidence accumulation, simulated firing rates/depolarization, and band-pass filtered local field potentials between 4 Hz and 32 Hz; no empirical dataset is analyzed.
Limitations
- Stated: the neural networks “should not be taken too seriously” and are offered as a starting point for refinement and deconstruction; many anatomical issues are ignored, including direct/indirect pathways and dopamine modulation.
- Stated: the sensory attenuation model is “quite a crude way” of modeling sensory attenuation, and the reading simulation is “a bit abstract.”
- Evident: the anatomical mappings are speculative and not empirically tested here, and the simulation results depend on the model assumptions and message-passing schedule.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures














Verification
Claims the paper makes
0 Valid · 1 Not valid · 45 Not checked · 146 Own results and interpretations
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Here, the subject correctly infers, on the last saccade of the last word, that the first sentence generated the stimuli.”
§ The graphical brain: Belief propagation and active inference > SIMULATIONS OF READING > Categorization and Evidence Accumulation
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This means that a partially observed sequence of outcomes can inform expectations about the future, which are necessary to evaluate the expected free energy of a policy.”
§ The graphical brain: Belief propagation and active inference > DISCRETE GENERATIVE MODELS > Belief Updating and Propagation
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
None of the sources cited for this claim have been checked yet.
In the paper
“When a generative model is specified as a factor graph, latent variables can often be inferred by executing a message passing schedule that can be derived automatically.”
§ The graphical brain: Belief propagation and active inference > INTRODUCTION
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A statement the paper makes itself — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Factor graphs are an important type of probabilistic graphical model because they facilitate the derivation of (approximate) Bayesian inference algorithms.”
§ The graphical brain: Belief propagation and active inference > INTRODUCTION
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Finally, note that the outcomes that inform the expected free energy are not the observed outcomes but predicted outcomes based upon expected states, under each policy (i.e., message 5).”
§ The graphical brain: Belief propagation and active inference > DISCRETE GENERATIVE MODELS
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
112 more claims
Create a free account to see all 192 claims and to verify the 73 references still open.
- Probably not a substantial claim. 39
- It may state two separate claims as one. 42
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“This means that systems able to make such future-oriented inferences must possess a generative model with temporal or counterfactual depth [26,39].”
From The Markov blankets of life: autonomy, active inference and the free energy principle (2018) · cites this paper as [39]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“In some simulations, one can mix discrete and continuous state-space models by placing the former on top of the latter, to produce deep generative models that, through active inference, can be used to simulate many known aspects of computational anatomy and physiology in the brain [158].”
From The free energy principle made simpler but not too simple (2023) · cites this paper as [158]
The cited source says what the claim attributes to it.
In the citing paper
“In this sense, it is sometimes referred to as epistemic affordance [158].”
From The free energy principle made simpler but not too simple (2023) · cites this paper as [158]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The graphical brain: Belief propagation and active inference > MIXED MODELS
“E(t)m=−lnoτ,m−∫0TL(t)mdtL(t)m=lnP(õ(t)|ηm)−lnP(õ(t)|η).(12)”
§ The graphical brain: Belief propagation and active inference > MIXED MODELS
“This equation expresses the free energy E of competing outcome models as the prior surprise plus the log evidence for each outcome model (integrated over time).”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The graphical brain: Belief propagation and active inference > SIMULATIONS OF READING > Categorization and Evidence Accumulation
“This figure illustrates belief propagation (producing the behavior shown in Figure 11), in terms of simulated firing rates and depolarization.”
§ The graphical brain: Belief propagation and active inference > SIMULATIONS OF READING
“Figure 12 shows the behavior that results from integrating the message passing scheme described in Figure 10.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The graphical brain: Belief propagation and active inference > SIMULATIONS OF READING
“each of the 25 time steps of the continuous integration scheme corresponds roughly to 10 ms”
§ The graphical brain: Belief propagation and active inference > SIMULATIONS OF READING
“This trajectory is shown as a function of the time steps used to integrate the generalized belief updating (24 time steps, each corresponding to about 10 ms of real time).”
Citations
Cites
(1988) · Cited by 16,582
Distributed hierarchical processing in the primate cerebral cortex
Felleman, D. J., & Van Essen, D. C. (1991) · Cerebral Cortex · doi:10.1093/cercor/1.1.1 · Cited by 7,783 · cited 2×
“Given that the message passing requires prediction errors on hidden causes to be passed to higher levels, one can assume they are encoded by superficial pyramidal cells; that is, the source of ascending extrinsic connections (Bastos et al., 2012; Felleman & Van Essen, 1991; Markov et al., 2013).”
BackgroundSelf-organized criticality: An explanation of the 1/f noise
Bak, P., Tang, C., & Wiesenfeld, K. (1987) · Physical Review Letters · doi:10.1103/physrevlett.59.381 · Cited by 7,591 · cited 1×
“From the perspective of active inference, this implies self-evidencing (Hohwy, 2016), which entails self-organized criticality (Bak et al., 1987) and dynamic connectivity (Allen et al., 2012; Breakspear, 2004).”
BackgroundFactor graphs and the sum-product algorithm
Kschischang, F. R., Frey, B. J., & Loeliger, H.-A. (2001) · IEEE Transactions on Information Theory · doi:10.1109/18.910572 · Cited by 6,486
Whatever next? Predictive brains, situated agents, and the future of cognitive science
Clark, A. (2013) · Behavioral and Brain Sciences · doi:10.1017/s0140525x12000477 · Cited by 5,924 · cited 1×
“In particular, we assume that the brain engages in some form of (Bayesian) inference—and can therefore be described as maximizing Bayesian model evidence (Clark, 2013; Friston, Kilner, & Harrison, 2006; Hohwy, 2016; Mumford, 1992).”
BackgroundRao, R. P. N., & Ballard, D. H. (1999) · Nature Neuroscience · doi:10.1038/4580 · Cited by 5,483 · cited 2×
“Before considering the implications for hierarchical architectures in the brain, we turn to the equivalent message passing for continuous variables, which transpires to be predictive coding (Rao & Ballard, 1999; Srinivasan, Laughlin, & Dubs, 1982).”
BackgroundA mechanism for cognitive dynamics: neuronal communication through neuronal coherence.
Fries, P. (2005) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2005.08.011 · Cited by 4,284 · cited 1×
“…dynamics) mechanism may be particularly important in understanding attention in terms of communication through coherence (Akam & Kullman, 2012; Fries, 2005) and the important role of inhibitory interneurons in mediating synchronous gain control (Kann, Papageorgiou, & Draguhn, 2014; Lee,…”
BackgroundHoeting, J. A., Madigan, D., Raftery, A. E., & Volinsky, C. T. (1999) · Statistical Science · doi:10.1214/ss/1009212519 · Cited by 4,118 · cited 1×
“See K. Friston and Penny (2011) and Hoeting et al. (1999) for details.”
BackgroundPreface: Cerebral Cortex Has Come of Age
Goldman-Rakic, P. S., & Rakic, P. (1991) · Cerebral Cortex · doi:10.1093/cercor/1.1.1-a · Cited by 3,393
Tracking whole-brain connectivity dynamics in the resting state
Allen, E. A., Damaraju, E., Plis, S. M., Erhardt, E. B., Eichele, T., & Calhoun, V. D. (2014) · Cerebral Cortex · doi:10.1093/cercor/bhs352 · Cited by 2,992 · cited 3×
“We focus on the modulation of intrinsic excitability; specifically, the afferents to superficial pyramidal cells—and, as a more abstract level, the implications for self-organized criticality of the sort entailed by dynamic fluctuations in connectivity (Aertsen, Gerstein, Habib, & Palm, 1989; Allen et al., 2012; Baker…”
Research works printed but not linked: 4
- Beal M. J. (2003). Variational algorithms for approximate Bayesian inference (Doctoral thesis, University College London).
- Friston K. J., Parr T., & de Vries B. (2017). Appendices for “The graphical brain: Belief propagation and active inference.” Network Neuroscience. 10.1162/netn_a_00018 PMC5798592 29417960
- Loeliger H.-A. (2002). Least squares and Kalman filtering on Forney graphs. In Blahut R. E. & Koetter R. (Eds.), Codes, graphs, and systems: A celebration of the life and career of G. David Forney, Jr., on the occasion of his sixtieth birthday (pp. 113– 135). Boston, MA: Springer US.
- MacKay D. J. C. (2003). Information theory inference and learning algorithms. Cambridge: Cambridge University Press.
Cited by
Items 201–300 of 408
Variational neuroethology: Answering further questions
Ramstead, M. J. D., Badcock, P. B., & Friston, K. J. (2018) · Physics of Life Reviews · doi:10.1016/j.plrev.2018.01.003 · Cited by 20
Self-citationA Computational Neuroscience Perspective on the Change Process in Psychotherapy
Smith, R. P., Lane, R. D., Nadel, L., & Moutoussis, M. (2020) · Oxford University Press eBooks · doi:10.1093/oso/9780190881511.003.0015 · Cited by 15
Editorial: Credition—An interdisciplinary approach to the nature of beliefs and believing
Seitz, R. J., Angel, H., Paloutzian, R. F., & Taves, A. (2023) · Frontiers in Behavioral Neuroscience · doi:10.3389/fnbeh.2023.1217648 · Cited by 15
“Furthermore, a post-hoc attribution is hardly compatible with a general neuroscientifically grounded model of belief formation and updating, as realized in parallel organized cortico-subcortical networks affording predictive processing (Friston et al., 2017; Seitz et al., 2018).”
BackgroundScience-Driven Societal Transformation, Part I: Worldview
(2020) · doi:10.3390/su12176881 · Cited by 13
Priors and prejudice: hierarchical predictive processing in intergroup perception
McGovern, H. T., & Otten, M. (2024) · Frontiers in Psychology · doi:10.3389/fpsyg.2024.1386370 · Cited by 16
“In more formal terms, this process is described as Bayesian belief updating, where message passing between neurons consists of an ongoing reciprocity of matching top-down predictions with bottom-up prediction errors (Friston et al., 2017).”
BackgroundCredition and the neurobiology of belief: the brain function in believing
Angel, H., & Seitz, R. J. (2024) · Academia Biology · doi:10.20935/acadbiol7359 · Cited by 14
Bayesian mechanics of perceptual inference and motor control in the brain
Kim, C. S. (2021) · Biological Cybernetics · doi:10.1007/s00422-021-00859-9 · Cited by 12
“Finally, we mention the recent research efforts on synthesizing perception, motor control, and decisionmakingwithin the FEP (Friston et al. 2015, 2017; Biehl et al. 2018; Parr and Friston 2019; van de Laar and de Vries 2019; Tschantz et al. 2020; Da Costa et al. 2020a).”
MethodParadoxical lesions, plasticity and active inference
Sajid, N., Parr, T., Gajardo‐Vidal, A., Price, C. J., & Friston, K. J. (2020) · Brain Communications · doi:10.1093/braincomms/fcaa164 · Cited by 12
Self-citationRealizing Active Inference in Variational Message Passing: The Outcome-Blind Certainty Seeker
Champion, T., Grześ, M., & Bowman, H. (2021) · Neural Computation · doi:10.1162/neco_a_01422 · Cited by 16
Affordance and Active Inference
Friston, K. J. (2022) · Springer eBooks · doi:10.1007/978-3-031-08629-8_20 · Cited by 18
Self-citationBayesian theories of consciousness: a review in search for a minimal unifying model
Rorot, W. (2021) · Neuroscience of Consciousness · doi:10.1093/nc/niab038 · Cited by 23
“…in continuous, discrete, andmixed-state spaces, with somewhat similar goals to the reinforcement learning approach in machine learning [see e.g. Sajid et al. (2021) for a review of applications in discrete cases and Friston et al. (2017b) for a discussion of the continuous and mixed cases].”
BackgroundAbove and beyond “Above and beyond the concrete”
Gilead, M., Trope, Y., & Liberman, N. (2020) · Behavioral and Brain Sciences · doi:10.1017/s0140525x20000023 · Cited by 14
Circular inference predicts nonuniform overactivation and dysconnectivity in brain-wide connectomes
Bouttier, V., Duttagupta, S., Denève, S., & Jardri, R. (2021) · Schizophrenia Research · doi:10.1016/j.schres.2020.12.045 · Cited by 14
Branching Time Active Inference: The theory and its generality
Champion, T., Costa, L. D., Bowman, H., & Grześ, M. (2022) · Neural Networks · doi:10.1016/j.neunet.2022.03.036 · Cited by 24
Are Generative Models Structural Representations?
Facchin, M. (2021) · Minds and Machines · doi:10.1007/s11023-021-09559-6 · Cited by 13
“…to capture the message passing within the brain; that is, how inference is performed (see, for instance, de Vries & Friston, 2017; Friston, 2017b; Friston, 2017c; Donnarumma, 2017; Matsumoto & Tani, 2020; see also Hinton & Sejnowski, 1983 for an earlier model).15 I believe that this is an…”
BackgroundInfluentialEdwards, D. J. (2025) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2025.1551960 · Cited by 13
Auzmendi, J., Moffatt, L., & Ramos, A. J. (2019) · Molecular Neurobiology · doi:10.1007/s12035-019-01749-9 · Cited by 10
“The construction of a Bayesian computational model (BCM) makes it possible to compare alternative theories regarding their ability to predict experimental data (see for example [8 – 10]).”
BackgroundBrain-Inspired Machine Intelligence: A Survey of Neurobiologically-Plausible Credit AssignmentPreprint
Ororbia, A. G. (2023) · OSF Preprints (OSF Preprints) · doi:10.31219/osf.io/xukqf · Cited by 37
Simulating lesion-dependent functional recovery mechanisms
Sajid, N., Holmes, E., Hope, T. M. H., Fountas, Z., Price, C. J., & Friston, K. J. (2021) · Scientific Reports · doi:10.1038/s41598-021-87005-4 · Cited by 9
Self-citationDeane, G. (2022) · Artificial Life · doi:10.1162/artl_a_00368 · Cited by 9
“As such, variational inference allows the posterior to converge to the “true” posterior through selecting beliefs that are associated with the minimum variational free energy (Friston, Parr, & de Vries, 2017).”
BackgroundActive Vision in Binocular Depth Estimation: A Top-Down Perspective
Priorelli, M., Pezzulo, G., & Stoianov, I. P. (2023) · Biomimetics · doi:10.3390/biomimetics8050445 · Cited by 9
A Survey on Cyber-Physical Security of Autonomous Vehicles Using a Context Awareness Method
Zaboli, A., Hong, J., Kwon, J., & Moore, J. (2023) · IEEE Access · doi:10.1109/access.2023.3338156 · Cited by 12
“The framework views these psychological/cognitive procedures and their interconnections as complementary modes of inference [31], [32], [33], [34].”
BackgroundDynamic planning in hierarchical active inference
Priorelli, M., & Stoianov, I. P. (2025) · Neural Networks · doi:10.1016/j.neunet.2024.107075 · Cited by 11
Systems of Bounded Rational Agents with Information-Theoretic Constraints
Gottwald, S., & Braun, D. A. (2018) · Neural Computation · doi:10.1162/neco_a_01153 · Cited by 11
“…to understand brain function in terms of active inference (Friston 2009, 2010; Friston, Levin et al., 2015; Friston, Rigoli et al., 2015; Friston, Lin et al., 2017; Friston, Parr et al., 2017), where perception is modeled as variational Bayesian inference over hidden causes of observations.”
BackgroundMethodInfluentialBranching time active inference: Empirical study and complexity class analysis
Champion, T., Bowman, H., & Grześ, M. (2022) · Neural Networks · doi:10.1016/j.neunet.2022.05.010 · Cited by 17
Tietz, S., Wagner‐Skacel, J., Angel, H., Ratzenhofer, M., Fellendorf, F. T., Fleischmann, E., Körner, C., Reininghaus, E. Z., Seitz, R. J., & Dalkner, N. (2022) · World Journal of Psychiatry · doi:10.5498/wjp.v12.i7.929 · Cited by 9
Events, Event Prediction, and Predictive Processing
Hohwy, J., Hebblewhite, A., & Drummond, T. W. (2020) · Topics in Cognitive Science · doi:10.1111/tops.12491 · Cited by 10
Between pebbles and organisms: weaving autonomy into the Markov blanket
van Es, T., & Kirchhoff, M. D. (2021) · Synthese · doi:10.1007/s11229-021-03084-w · Cited by 13
“…in cells (Auletta, 2013; Friston, 2013), neuronal signalling in brains (Friston et al., 2017; Parr & Friston, 2017), tropism in plants (Calvo & Friston, 2017), synchronised singing in birds (Friston & Frith, 2015) to decision-making and planning in mammals (Daunizeau et al., 2010; Friston,…”
BackgroundSmith, R., Kirlic, N., Stewart, J. L., Touthang, J., Kuplicki, R., Khalsa, S. S., Paulus, M. P., Investigators, T., & Aupperle, R. (2020) · doi:10.31234/osf.io/t2dhn · Cited by 10
How mood shapes belief updating bias in depression
Bottemanne, H., Frileux, S., Sévoz‐Couche, C., Pelloux, Y., Colle, R., & Corruble, E. (2025) · Cognitive Affective & Behavioral Neuroscience · doi:10.3758/s13415-025-01297-x · Cited by 12
Depth and hierarchies in the predictive brain: From reaction to action
Muzik, O., & Diwadkar, V. A. (2023) · Wiley Interdisciplinary Reviews Cognitive Science · doi:10.1002/wcs.1664 · Cited by 8
A Goal-Directed Trajectory Planning Using Active Inference in UAV-Assisted Wireless Networks
Krayani, A., Khan, K., Marcenaro, L., Marchese, M., & Regazzoni, C. S. (2023) · Sensors · doi:10.3390/s23156873 · Cited by 10
“Probabilistic graphical models are used to represent active inference models because they provide a clear visual representation of the model’s computational structure and how belief updates can be achieved through message-passing algorithms [37].”
MethodIntelligent Resource Allocation for UAV-Based Cognitive NOMA Networks: An Active Inference Approach
Obite, F., Krayani, A., Alam, A. S., Marcenaro, L., Nallanathan, A., & Regazzoni, C. S. (2023) · IEEE Future Networks World Forum (FNWF) · doi:10.1109/fnwf58287.2023.10520397 · Cited by 8
“Conversely, active inference, an emerging approach from theoretical neuroscience, offers a comprehensive brain theory that unifies action, perception, and inference for adaptive systems [7].”
BackgroundParvizi-Wayne, D., & Severs, L. (2024) · Cognitive Affective & Behavioral Neuroscience · doi:10.3758/s13415-024-01189-6 · Cited by 9
“…a corollary of the Free Energy Principle, which holds that any thing that persists through time—in so far as it frequents a limited set of characteristic states— attains a nonequilibrium steady state (Friston, 2009, 2010, 2019; Friston et al., 2017a; Kirchhoff et al., 2018; Ramstead et al., 2023a).”
BackgroundBrahinsky, J., Mago, J., Miller, M. M., Catherine, S., & Lifshitz, M. (2024) · American Journal of Human Biology · doi:10.1002/ajhb.24189 · Cited by 11
Learning cognitive maps as structured graphs for vicarious evaluationPreprint
Rikhye, R. V., Gothoskar, N., Guntupalli, J. S., Dedieu, A., Lázaro-Gredilla, M., & George, D. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/864421 · Cited by 10
Natural selection finds natural gradientPreprint
Costa, L. D., Parr, T., Sengupta, B., & Friston, K. J. (2020) · arXiv (Cornell University) · Cited by 10
Self-citationNatural language syntax complies with the free-energy principlePreprint
Murphy, E., Holmes, E., & Friston, K. J. (2021) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/r3fcx · Cited by 8
Self-citationRepresentation learning with reward prediction errors
Alexander, W. H., & Gershman, S. J. (2022) · Neurons Behavior Data analysis and Theory · doi:10.51628/001c.37270 · Cited by 11
Experienced meditators show greater forward traveling cortical alpha wave strengths
Bailey, N. W., Hill, A. T., Godfrey, K., Perera, M. P. N., Hohwy, J., Corcoran, A. W., Rogasch, N. C., Fitzgibbon, B. M., & Fitzgerald, P. B. (2025) · Annals of the New York Academy of Sciences · doi:10.1111/nyas.15401 · Cited by 7
Inferential planning in the frontal cortex
Donnarumma, F., Parr, T., Friston, K. J., Whittington, J. C. R., & Pezzulo, G. (2026) · Cell Reports · doi:10.1016/j.celrep.2026.117905 · Cited by 7
Self-citationSlow but flexible or fast but rigid? Discrete and continuous processes compared
Priorelli, M., & Stoianov, I. P. (2024) · Heliyon · doi:10.1016/j.heliyon.2024.e39129 · Cited by 9
“This approach has not received adequate attention yet, as not many implementations can be found in the literature [2, 18, 34, 38, 35, 39, 40], none simulating applications in dynamic contexts.”
BackgroundInfluentialFrAMBI: A Software Framework for Auditory Modeling Based on Bayesian Inference
Barumerli, R., & Majdak, P. (2025) · Neuroinformatics · doi:10.1007/s12021-024-09702-5 · Cited by 9
“This systema-tisation increases reproducibility (Riquelme & Gjorgjieva, 2021) since it can effectively reuse methods and models across studies (Friston et al., 2017).”
Priors and proprioceptive predictions
Parr, T., Ramstead, M. J. D., & Friston, K. J. (2025) · Current Opinion in Behavioral Sciences · doi:10.1016/j.cobeha.2025.101509 · Cited by 13
“There is an emerging view, rooted in the concept of motor chunking [53], that these predictions should anticipate sequences of short motor trajectories that can be combined to produce complex movements [54].”
Self-citationChance-Constrained Active Inference
van de Laar, T., Şenöz, İ., Özçelikkale, A., & Wymeersch, H. (2021) · Neural Computation · doi:10.1162/neco_a_01427 · Cited by 11
“, 2004), and can thus be used to complement message-passing formulations on generative neural models (Friston et al., 2017; van de Laar, Cox, Senoz, Bocharov, & de Vries, 2018).”
BackgroundMethodAgentive Cognitive Construction Grammar: a predictive semiotic theory of mind and language
Torres–Martínez, S. (2024) · Semiotica · doi:10.1515/sem-2018-0138 · Cited by 9
Dynamic Inference by Model Reduction
Priorelli, M., & Stoianov, I. P. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3595670 · Cited by 7
“This approach can be used, e.g., to reach fixed positions in sequence [1] or simulate pictographic reading [25].”
Unification of MAP Estimation and Marginal Inference in Recurrent Neural Networks
Yu, Z., Chen, F., & Deng, F. (2018) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2018.2805813 · Cited by 7
“In this context, variational formulations of BP speak to the biological plausibility of recurrent neural networks (RNNs) as for real neuronal processing [18].”
Peraza, J. A., Martínez-Montes, E., Aubert, E., Valdes-Hernandez, P. A., & Mulet, R. (2019) · Neural Networks · doi:10.1016/j.neunet.2019.11.014 · Cited by 9
A variational approach to scriptsPreprint
Albarracin, M., Constant, A., Friston, K. J., & Ramstead, M. J. D. (2020) · OSF Preprints (OSF Preprints) · doi:10.31219/osf.io/x45ha · Cited by 7
Self-citationVisual mismatch negativity in Parkinson's psychosis and potential for testing treatment mechanisms
Vignando, M., ffytche, D., Mazibuko, N., Palma, G., Montagnese, M., Dave, S., Nutt, D. J., Gabay, A. S., Tai, Y. F., Batzu, L., et al. (2024) · Brain Communications · doi:10.1093/braincomms/fcae291 · Cited by 6
“61 A more in-depth analysis of models of VH in PD has been recently carried out to create a consensus framework for VH in Parkinson’s.”
BackgroundThe role of affective states in computational psychiatry
Benrimoh, D. A., Smith, R. S., Diaconescu, A. O., Friesen, T., Jalali, S., Mikuš, N., Gschwandtner, L. P., Gandhi, J., Horga, G., & Powers, A. R. (2025) · The International Journal of Neuropsychopharmacology · doi:10.1093/ijnp/pyaf049 · Cited by 8
“…regarding how patterns of connections within idealized cortical columns could plausibly implement the update equations in active inference, 71 no detailed hypotheses have been tested to date regarding possible micro-or macro-scale realizations in the brain (eg, more biologically detailed…”
Gregoretti, F., Pezzulo, G., & Maisto, D. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.127065 · Cited by 7
Cortical development in the structural model and free energy minimization
Wright, J., & Bourke, P. D. (2024) · Cerebral Cortex · doi:10.1093/cercor/bhae416 · Cited by 6
Self-regulation learning as active inference: dynamic causal modeling of an fMRI neurofeedback task
Vargas, G. S., Araya, D., Sepúlveda, P., Rodriguez-Fernández, M., Friston, K. J., Sitaram, R., & El‐Deredy, W. (2023) · Frontiers in Neuroscience · doi:10.3389/fnins.2023.1212549 · Cited by 7
“Indeed, Active inference offers a reinterpretation (Pezzulo et al., 2015; Friston et al., 2017), where a shift from striatal bottomup to Frontal top-down processing has been proposed as a better explanation of learning (Smith et al.”
BackgroundSelf-citationScience-Driven Societal Transformation, Part III: Design
Boik, J. C. (2021) · Sustainability · doi:10.3390/su13020726 · Cited by 9
“This is in nod to the work of Karl Friston, the neuroscientist who is primarily responsible for the concept of active inference [42,43].”
BackgroundWSN-Driven Posture Recognition and Correction Towards Basketball Exercise
Cai, X. (2022) · International Journal of Information System Modeling and Design · doi:10.4018/ijismd.300777 · Cited by 6
“Belief propagation (BP) (Friston et al. 2017) is a commonly used graph model optimization algorithm.”
Wiest, M. C., & Puniani, A. S. (2025) · Computational and Structural Biotechnology Journal · doi:10.1016/j.csbj.2025.09.017 · Cited by 6
“Several process models are able to predict neuronal activity in specific stimulus contexts [99,30,41,48,92] or accompanying specific behaviors [99,103,105,2,47,46,50,96,95].”
Joint modeling of choices and reaction times based on Bayesian contextual behavioral control
Schwöbel, S., Marković, D., Smolka, M. N., & Kiebel, S. J. (2024) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1012228 · Cited by 6
Harris, D., Arthur, T., Wilson, M., Gallais, B. L., Parsons, T. D., Dill, A., & Vine, S. (2025) · Biological Cybernetics · doi:10.1007/s00422-025-01006-4 · Cited by 6
Horizontal connectivity in V1: Prediction of coherence in contour and motion integration
Bec, B. L., Troncoso, X. G., Desbois, C., Passarelli, Y., Baudot, P., Monier, C., Pananceau, M., & Frégnac, Y. (2022) · PLoS ONE · doi:10.1371/journal.pone.0268351 · Cited by 5
Safron, A. (2023) · Communications in computer and information science · doi:10.1007/978-3-031-28719-0_18 · Cited by 6
Beliefs Made It into Science: Believe It or Not
Seitz, R. J., & Paloutzian, R. F. (2023) · Function · doi:10.1093/function/zqad049 · Cited by 7
Yin, H. H. (2025) · Current Opinion in Behavioral Sciences · doi:10.1016/j.cobeha.2025.101487 · Cited by 5
SR-AIF: Solving Sparse-Reward Robotic Tasks From Pixels with Active Inference and World Models
Nguyen, V. D., Yang, Z., Buckley, C. L., & Ororbia, A. G. (2025) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra55743.2025.11127713 · Cited by 14
“In active inference, the agent estimates both future states and observations and then plans action sequences based on the expected free energy computed from these future ‘imagined realities’ [24, 19, 18, 14].”
BackgroundMemetics and neural models of conspiracy theories
Duch, W. (2021) · Patterns · doi:10.1016/j.patter.2021.100353 · Cited by 8
“Bayesian models of belief propagation are used to model details of perceptual processes and relate them to connectomes.(2) The artificial neural network community has focused on faithful learningmethods, but there is another, neglected side of learning and memory formation.”
MethodRemembrance of things perceived: Adding thalamocortical function to artificial neural networks
Loeb, G. E. (2023) · Frontiers in Integrative Neuroscience · doi:10.3389/fnint.2023.1108271 · Cited by 6
“The thalamocortical system has been proposed as the substrate for the generation of predictions of sensory information based on such probabilities and their comparison with actual sensory signals (Friston et al., 2017a).”
BackgroundPassive motion and active inference
Friston, K. J., & Parr, T. (2019) · Physics of Life Reviews · doi:10.1016/j.plrev.2019.01.004 · Cited by 6
Self-citationThe 2-D Cluster Variation Method: Topography Illustrations and Their Enthalpy Parameter Correlations
Maren, A. J. (2021) · Entropy · doi:10.3390/e23030319 · Cited by 7
“(2020) addresses susceptibility propagation (SP), belief propagation (BP), and critical state dynamics [25], and Friston, Parr, and de Vries (2017) investigate the connection between belief propagation and active inference [26].”
BackgroundUnderstanding the Origin of Information-Seeking Exploration in Probabilistic Objectives for ControlPreprint
Millidge, B., Seth, A. C., & Buckley, C. L. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.06859 · Cited by 18
Kalman filters as the steady-state solution of gradient descent on variational free energyPreprint
Baltieri, M., & Isomura, T. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.10530 · Cited by 12
“These include, among others, predictive coding [82, 83], optimal feedback control [88, 87] and active inference [45, 41].”
BackgroundA variational approach to scriptsPreprint
Albarracin, M., Constant, A., Friston, K. J., & Ramstead, M. J. D. (2020) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/67zy4 · Cited by 8
Self-citationPossible Principles for Aligned Structure Learning Agents
Costa, L. D., Gavenčiak, T., Hyland, D., Samiei, M., Dragos-Manta, C., Pattisapu, C., Razi, A., & Friston, K. J. (2025) · Neural Computation · doi:10.1162/neco.a.39 · Cited by 8
Self-citationPrest, S. (2026) · Neural Computation · doi:10.1162/neco.a.1534 · Cited by 5
On Predictive Planning and Counterfactual Learning in Active Inference
Paul, A., Isomura, T., & Razi, A. (2024) · Entropy · doi:10.3390/e26060484 · Cited by 11
“Given this general foundation, active inference [6] offers excellent flexibility in defining the generative model structure for a given problem, and it has attracted much attention in various domains [7,8].”
BackgroundParr, T., Oswal, A., & Manohar, S. (2024) · Neuroscience & Biobehavioral Reviews · doi:10.1016/j.neubiorev.2024.105984 · Cited by 7
Self-citationLearning in Hybrid Active Inference Models
Collis, P., Singh, R. R., Kinghorn, P., & Buckley, C. L. (2024) · Communications in computer and information science · doi:10.1007/978-3-031-77138-5_4 · Cited by 8
The Emperor Is Naked: Replies to commentaries on the target article
Bruineberg, J. P., Dołęga, K., Dewhurst, J., & Baltieri, M. (2022) · Behavioral and Brain Sciences · doi:10.1017/s0140525x22000656 · Cited by 6
Thinking like a State : Embodied intelligence in the deep history of our collective mind
GUÉNIN--CARLUT, A. (2022) · IOP Conference Series Materials Science and Engineering · doi:10.1088/1757-899x/1261/1/012026 · Cited by 8
“While the enactive approach grounds its study of cognition in the way an organism enacts its structural identity, Active Inference formalises the relation between the structural architecture of a cognitive system (ie what constraints define it) (9) and its functional architecture (ie what world it enacts) (85).”
BackgroundCardon, G. J., Buckhannon, M., & Rojas, D. C. (2023) · Biological Psychology · doi:10.1016/j.biopsycho.2023.108712 · Cited by 6
Beliefs in Pain and Suffering: A Cognitive Neuroscience Approach
Seitz, R. J. (2024) · Scientia et Fides · doi:10.12775/setf.2024.003 · Cited by 7
“Information processing is mediated in parallel cortico-subcortical loops in accordance with the free-energy principle as proposed by Friston et al. (2017).”
MethodFriston, K. J. (2020) · Behavioral and Brain Sciences · doi:10.1017/s0140525x19002929 · Cited by 5
Self-citationWaade, P. T., Mikuš, N., & Mathys, C. D. (2021) · Communications in computer and information science · doi:10.1007/978-3-030-93736-2_57 · Cited by 7
Teleosemantics and the free energy principle
Mann, S. F., & Pain, R. (2022) · Biology & Philosophy · doi:10.1007/s10539-022-09868-9 · Cited by 5
Integrated World Modeling Theory (IWMT) RevisitedPreprint
Safron, A. (2019) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/kjngh · Cited by 4
A mixed generative model of auditory word repetitionPreprint
Sajid, N., Holmes, E., Costa, L. D., Price, C. J., & Friston, K. J. (2022) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2022.01.20.477138 · Cited by 6
“To do this, we cast word repetition as a process of active inference (Friston et al., 2017c; Da Costa et al., 2020a; Friston et al., 2020b), which treats belief updating as a gradient descent on variational free energy (Hinton and Zemel, 1993), analogous to the evidence lower bound (Winn and…”
BackgroundMethodInfluentialSelf-citationConstant, A., Albarracin, M., Perin, M., Thiruvengada, H., & Friston, K. J. (2025) · Frontiers in Sustainable Cities · doi:10.3389/frsc.2025.1571613 · Cited by 6
Self-citationImproving the Minimum Free Energy Principle to the Maximum Information Efficiency Principle
Lu, C. (2025) · Entropy · doi:10.3390/e27070684 · Cited by 4
“Friston et al. first applied VB to brain science and later extended it to biobehavioral science, thus developing the MFE criterion into the FEP [6,7].”
Neural mechanisms of emotion-focused interventions: A meta-analytic review of fMRI studies
Li, Y., Li, G., Liu, Y., Liu, C., & Chen, A. (2025) · NeuroImage · doi:10.1016/j.neuroimage.2025.121469 · Cited by 4
A case for chaos theory inclusion in neuropsychoanalytic modeling
Tolchinsky, A. (2023) · Neuropsychoanalysis · doi:10.1080/15294145.2023.2191983 · Cited by 4
“Indeed, it is not uncommon to suppose that the brain might use both at the same time; namely, continuous state space models to process a continuous time engagement with the world, which is modeled in terms of discrete representations that are more apt to handle things like concepts and language (Friston et al., 2017).”
BackgroundInterpersonal strategy for controlling unpredictable opponents in soft tennis
Yamamoto, Y., Yokoyama, K., Kijima, A., Okumura, M., Shima, H., & Gohara, K. (2024) · Scientific Reports · doi:10.1038/s41598-024-71538-5 · Cited by 5
“This strategy is believed to involve skilled players capitalizing on their opponents’ intentions to predict or reduce uncertainty 11,22 .”
BackgroundA detailed theory of thalamic and cortical microcircuits for predictive visual inference
George, D., Lázaro-Gredilla, M., Lehrach, W. P., Dedieu, A., Zhou, G., & Marino, J. (2025) · Science Advances · doi:10.1126/sciadv.adr6698 · Cited by 13
Deep Hybrid Models: Infer and Plan in a Dynamic World
Priorelli, M., & Stoianov, I. P. (2025) · Entropy · doi:10.3390/e27060570 · Cited by 5
“Considering the example in Figure 3.1.1, the hidden states x and x ′ may encode the angle and angular velocity of an agent’s arm, and two hidden causes v circle and v square may be associated with dynamics functions f circle and f square encoding (potential) reaching movements toward the two…”
InfluentialSynthetic Spatial Foraging With Active Inference in a Geocaching Task
Neacsu, V., Convertino, L., & Friston, K. J. (2022) · Frontiers in Neuroscience · doi:10.3389/fnins.2022.802396 · Cited by 6
“In what follows, we summarize active inference with its implicit minimization of variational and expected free energy, offering a brief account of planning as inference (Botvinick and Toussaint, 2012; Friston et al., 2017a).”
BackgroundMethodInfluentialSelf-citationFlights and Perchings of the BrainMind: A Temporospatial Approach to Psychotherapy
Chan, A., Northoff, G., Karasik, R., Ouyang, J., & Williams, K. (2022) · Frontiers in Psychology · doi:10.3389/fpsyg.2022.828035 · Cited by 3
“Models of Bayesian inference have exhibited temporal nestedness in which state transitions proceed at different rates at different levels of the hierarchy, generating outcomes over different, nested timescales (Friston et al., 2017).”
Branching Time Active Inference with Bayesian Filtering
Champion, T., Grześ, M., & Bowman, H. (2022) · Neural Computation · doi:10.1162/neco_a_01529 · Cited by 4
“The first considers a policy that starts at the beginning of a trial (Friston, Parr, and de Vries, 2017), that is, zero-tethered active inference.”
BackgroundA Meta-Bayesian Model of Intentional Visual SearchPreprint
Cullen, M., Monney, J., Mirza, M. B., & Moran, R. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.03531 · Cited by 4
“This issue is addressed by supplementing the agent’s generative model with an additional (subordinate) continuous-time model that can accumulate evidence from the visual domain [29] i.”
BackgroundThinking like a State - Embodied intelligence in the deep history of our collective mindsPreprint
GUÉNIN--CARLUT, A. (2021) · OSF Preprints (OSF Preprints) · doi:10.31219/osf.io/dxnzt · Cited by 4
Leptourgos, P. (2018) · theses.fr (ABES) · doi:10.70675/dafacdb1z9072z4ec2z8cebz0ca09b93b7ce · Cited by 4
Metadata
- Authors
- Friston, K. J., Parr, T., & de Vries, B.
- Affiliations
- Wellcome Centre for Human Neuroimaging (GB)
- Eindhoven University of Technology, Department of Electrical Engineering, Eindhoven, The Netherlands
- Corresponding author
- Parr, Thomas
- Venue
- Network Neuroscience
- Publisher
- MIT Press
- Published
- 2017-12
- Volume
- 1
- Issue
- 4
- Pages
- 381-414
- ISSN
- 2472-1751 (ISSN-L)
- PMC
- PMC5798592
- OpenAlex
- W2623066203
- Cited by
- 415
- References
- 111
- Citations by year
- 2026: 49 · 2025: 39 · 2024: 41 · 2023: 61 · 2022: 51 · 2021: 56 · 2020: 55 · 2019: 32 · 2018: 25 · 2017: 6
- FWCI
- 18.51
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- cc-by
- Locations
- Network Neuroscience (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- TU/e Research Portal (submittedVersion, other-oa)
- TU/e Research Portal (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- UCL Discovery (University College London) (submittedVersion)
- TU/e Research Portal (submittedVersion, cc-by)
- Funders
- Wellcome Trust (088130/Z/09/Z)
- Rosetrees Trust (173346)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
- Physical Sciences › Computer Science › Artificial Intelligence › Bayesian Modeling and Causal Inference
- Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
- Keywords
- active inference, belief propagation, deep generative models, factor graphs, Bayesian model averaging, thalamocortical loop, functional integration, reading, semantic representation, visual sampling
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.
A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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
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.