The DOI matches its Crossref record; 129 of the 130 references it cites were found in a registry or an index.
Neural Dynamics under Active Inference: Plausibility and Efficiency of Information Processing
Open accessAbstract
Active inference is a normative framework for explaining behaviour under the free energy principle-a theory of self-organisation originating in neuroscience. It specifies neuronal dynamics for state-estimation in terms of a descent on (variational) free energy-a measure of the fit between an internal (generative) model and sensory observations. The free energy gradient is a prediction error-plausibly encoded in the average membrane potentials of neuronal populations. Conversely, the expected probability of a state can be expressed in terms of neuronal firing rates. We show that this is consistent with current models of neuronal dynamics and establish face validity by synthesising plausible electrophysiological responses. We then show that these neuronal dynamics approximate natural gradient descent, a well-known optimisation algorithm from information geometry that follows the steepest descent of the objective in information space. We compare the information length of belief updating in both schemes, a measure of the distance travelled in information space that has a direct interpretation in terms of metabolic cost. We show that neural dynamics under active inference are metabolically efficient and suggest that neural representations in biological agents may evolve by approximating steepest descent in information space towards the point of optimal inference.
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.
Integrity signals
Registry
Cites
Statistics
Figures
![Figure 1 Active inference for state estimation (discrete state-space). Panel (a) summarises the problem of finding the minimum of a function (e.g., the free energy). One possibility would be taking the shortest path, which involves climbing up a hill; however, in the nescience of the minimum, a viable strategy consists of myopically taking the direction of steepest descent. In panel (b), we depict an example of a trajectory of an agent’s beliefs during the process of perception, which consists of updating beliefs about the states of the external world to reach the point of optimal inference (i.e., free energy minimum). In this example, the state-space comprises only three states (e.g., three different locations in a room). As they are probabilities over states, the components of s are non-negative and sum to one; hence, the agent’s beliefs naturally live on a triangle in three-dimensional space. Mathematically, this object is called a (two-dimensional) simplex. This constitutes the belief space, or information space, on which the free energy is defined. Technically, this object is a smooth statistical manifold, which corresponds to the set of parameters of a categorical distribution. To optimise metabolic and computational efficiency, agents must update their beliefs to reach the free energy minimum via the shortest possible path on this manifold. In panel (c), we exhibit simulated local field potentials that arise by interpreting the rate of change of v in terms of depolarisations, over a sequence of eight observations (e.g., saccadic eye-movements). As the rate of change is given by the free energy gradients, the decay of these local field potentials to zero coincides with the reaching of the free energy minimum (at which the gradient is zero by definition). Each colour expresses the mean voltage potential of a neural population, which encodes beliefs about one hidden state. These were obtained during the first eight trials of the T-Maze numerical simulation described in Section 6.1.1. Local field potentials accompanying the rule learning simulation—and for subsequent trials of the T-Maze task—can be found in [39,57], respectively. For more details on the generation of simulated electrophysiological responses, see [57].](https://figures.cochat.ai/paper-figures/8f38519301984776b6b1ec1b1dd72aa3/neural-dynamics-under-active-inference-plausibility-and-efficiency-of-figure-1.jpg)

![Figure 3 Generative model and T-Maze task. This figure describes the generative model used in the simulations and the specificities of the T-Maze task. The generative model is a partially observed Markov decision process (POMDP), a canonical model for sequential decision-making tasks under uncertainty. On the left-hand-side, the generative model is expressed as a factor graph [122]: the circles represent random variables while the squares represent factors. The shaded variables are observations, which the agent samples in discrete time. The unshaded variables are random variables that are inferred by the agent. These are either hidden states (s) or policies (π, i.e., action sequences), which need to be inferred from observations. The factors encode the relationships between the variables that they link. For example, D is a vector of non-negative components that sum to one that expresses the prior belief of being in this or another hidden state; B is a stochastic matrix that encodes the transition probabilities between hidden states given actions; similarly, A is a matrix that gives the probability of observations given states. See [7] for more details. Importantly, the factors can themselves be learned over time. On the right-hand side, we further unpack the generative model of the T-Maze task. The task is as pictured in the upper-right corner. The rat’s initial location is the centre of the T-Maze. At each time-step, the rat can choose between four control states (i.e., actions) that lead the rat to move to one of the four maze locations (middle, top-left, top-right or bottom). There are four hidden states that correspond to the location of the rat in the T-Maze and another two that correspond to the location of the reward (right arm or left arm), which leads to eight hidden states in total, meaning that the associated simplex on which the free energy is defined is seven-dimensional. There are seven possible observations that correspond to being in the middle of the T-Maze, in each one of the upper arms with (in red) or without the reward (in grey), and to collecting the cue, which reveals the location of the reward. The agent has a preference for each of these observations: the agent wants to claim the reward and avoid the unbaited arm. The agent is neutral w.r.t. all other observations. See [57] for more details of this paradigm.](https://figures.cochat.ai/paper-figures/d12deb176f9e4582b7e4394a3ea8c6e8/neural-dynamics-under-active-inference-plausibility-and-efficiency-of-figure-3.jpg)
![Figure 4 Information length of perceptual inference in active inference and natural gradient. The upper-half of the figure plots the results of the numerical experiments with T-Maze task, while the lower half the abstract rule learning task. In Panel (a), the histograms show the information length of perceptual inference in active inference (in red) and natural gradient (in blue). Specifically, this is (mean = 7075, median = 2.9776) for active inference, and (mean = 2.7793, median = 3.1163) for the natural gradient; in the rule learning task, the information lengths are (mean = 14.3573, median = 16.2099) for active inference and (mean = 13.8588, median = 16.1058) for the natural gradient. Information trajectories are longer in the abstract rule learning task, due to its increased complexity. In other words, the greater number of hidden states—in the generative model—mandate more belief updating. This means the model used for this high-dimensional task entails greater complexity in the technical sense: as belief updating moves posterior beliefs further from their priors. (Parenthetically, this means that the agents took much longer to update their beliefs). The reasons for systematic variation in information length across trials within tasks is that (1) the task configuration varied from trial to trial and (2) the generative models were themselves learned over trials. The sudden increase in information length in the T-Maze task at the ninth trial occurs when agents learn that the reward’s location is constant. As the reward’s location is unambiguous, belief trajectories are longer as beliefs about context move from the middle of the simplex to the vertex representing the reward’s location at each step of the paradigm. This means that we can improve the biological plausibility of the implementation by starting each belief update at the previous free energy minimum instead of an agnostic location in the middle of the simplex. This would avoid beliefs travelling to the same vertex multiple times in the same trial and hence would make belief trajectories shorter once the reward location is learned. This consideration does not affect our characterisation, which simply compares the information length given consistent starting and end points. As the context could be deduced only at the last step of the task, a similar sudden change in information length did not occur in the abstract rule learning task. However, the slight increase in information length across trials occurs for the same reason: learning the generative model made inferences increasingly precise. In Panel (b), the violin plots illustrate the difference in information length by subtracting the information length of the natural gradient from the information length of active inference. In the T-Maze paradigm, active inference mostly takes shorter paths, which is why the violin plot’s values are mostly negative (mean= −0.0718, median= −0.1146). However, we obtain the opposite pattern in the abstract rule learning task (mean = 0.4985, median= 0.5401). Panel (c) shows an example of the belief trajectory taken during state estimation in the abstract rule learning task. This was possible to include as the generative model includes a hidden state dimension with three possible alternatives [39], enabling a two-dimensional representation of the associated simplex. The red trajectory corresponds to active inference, the blue is natural gradient descent, and the orange is the shortest path to the free energy minimum (i.e., the geodesic, see Appendix D). This example is not representative of the average and was chosen for purely illustrative purposes as the trajectories are very distinct, lengthy and do not coincide with the geodesic. The fact that both schemes were significantly suboptimal in this example was unsurprising as beliefs evolve to the free energy minimum using only local information about the free energy landscape. Note that this suboptimality occurred only in a small minority of the trials considered here.](https://figures.cochat.ai/paper-figures/71f31cf9270f47c39b2e51a4ea042b1b/neural-dynamics-under-active-inference-plausibility-and-efficiency-of-figure-4.jpg)
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.
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 cited source says what the claim attributes to it.
In the citing paper
“The ensuing minimisation of free energy can be formulated as gradient flows following (17)—between the discrete arrival of new sensory input—in a way that relates comfortably to neuronal dynamics [123, 124, 42].”
From The free energy principle made simpler but not too simple (2023) · cites this paper as [42]
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
On Information and Sufficiency
Kullback, S., & Leibler, R. A. (1951) · The Annals of Mathematical Statistics · doi:10.1214/aoms/1177729694 · Cited by 20,193
Elements of Information Theory
Cover, T. M., & Thomas, J. A. (2005) · doi:10.1002/047174882x · Cited by 13,791
The free-energy principle: a unified brain theory?
Friston, K. (2010) · Nature Reviews Neuroscience · doi:10.1038/nrn2787 · Cited by 8,120 · cited 1×
“Active inference unifies many existing theories of brain function [13,14], such as, for example, optimal control [15–18], the Bayesian brain hypothesis [19–21] and predictive coding [19,22–24].”
MethodSelf-citationEmergence of simple-cell receptive field properties by learning a sparse code for natural images
Olshausen, B. A., & Field, D. J. (1996) · Nature · doi:10.1038/381607a0 · Cited by 5,694
Rao, R. P. N., & Ballard, D. H. (1999) · Nature Neuroscience · doi:10.1038/4580 · Cited by 5,483
Irreversibility and heat generation in the computing process
Landauer, R. (1961) · IBM Journal of Research and Development · doi:10.1147/rd.53.0183 · Cited by 4,980
Receptive fields of single neurones in the cat's striate cortex
Hubel, D. H., & Wiesel, T. N. (1959) · The Journal of Physiology · doi:10.1113/jphysiol.1959.sp006308 · Cited by 4,653 · cited 2×
“Most notably, the seminal work of Hubel and Wiesel [106] showed the existence of neurons encoding orientation of visual stimuli.”
BackgroundFriston, K. J., Harrison, L., & Penny, W. (2003) · NeuroImage · doi:10.1016/s1053-8119(03)00202-7 · Cited by 4,583 · cited 1×
“This important fact, which was verified experimentally using dynamic causal modelling [88,90], means that the variance of membrane potentials implicitly encodes the (inverse) precision of the information encoded within the population.”
BackgroundSelf-citationA theory of cortical responses.
Friston, K. (2005) · Philosophical Transactions of the Royal Society B: Biological Sciences · doi:10.1098/rstb.2005.1622 · Cited by 4,578
Self-citationMicrostructure of a spatial map in the entorhinal cortex
Hafting, T., Fyhn, M., Molden, S., Moser, M.-B., & Moser, E. I. (2005) · Nature · doi:10.1038/nature03721 · Cited by 4,068 · cited 1×
“This is the raison d’être of the study of place cells [99], grid cells [100] and head-direction cells [101,102], where the states inferred are physical locations in space.”
BackgroundAn Introduction to Variational Methods for Graphical Models
Jordan, M. I., Ghahramani, Z., Jaakkola, T. S., & Saul, L. K. (1999) · Machine Learning · doi:10.1023/a:1007665907178 · Cited by 3,791
Natural Gradient Works Efficiently in Learning
Amari, S.-I. (1998) · Neural Computation · doi:10.1162/089976698300017746 · Cited by 2,681 · cited 4×
“Amari’s natural gradient descent [85,116] is a well-known optimisation algorithm for finding the minimum of functions defined on statistical manifolds such as the variational free energy.”
BackgroundInfluentialThe Bayesian brain: The role of uncertainty in neural coding and computation
Knill, D. C., & Pouget, A. (2004) · Trends in Neurosciences · doi:10.1016/j.tins.2004.10.007 · Cited by 2,588
Building machines that learn and think like people
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2017) · Behavioral and Brain Sciences · doi:10.1017/s0140525x16001837 · Cited by 2,550
Building Machines That Learn and Think Like People
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., & Gershman, S. J. (2016) · arXiv · doi:10.48550/arxiv.1604.00289 · Cited by 2,550
Natural Image Statistics and Neural Representation
Simoncelli, E. P., & Olshausen, B. A. (2001) · Annual Review of Neuroscience · doi:10.1146/annurev.neuro.24.1.1193 · Cited by 2,460
Canonical Microcircuits for Predictive Coding
Bastos, A. M., Usrey, W. M., Adams, R. A., Mangun, G. R., Fries, P., & Friston, K. J. (2012) · Neuron · doi:10.1016/j.neuron.2012.10.038 · Cited by 2,421
Self-citation(1998) · Springer eBooks · doi:10.1007/978-94-011-5014-9 · Cited by 2,418
A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue
Wilson, H. R., & Cowan, J. D. (1973) · Kybernetik · doi:10.1007/bf00288786 · Cited by 1,850 · cited 2×
“The sigmoid relationship between membrane potential and firing rate, was originally derived by Wilson and Cowan [86], who showed that any unimodal distribution of thresholds within a neural population, whose individual neurons are modelled as a Heaviside response unit, results in a sigmoid activation function at the p…”
BackgroundThe free-energy principle: a rough guide to the brain?
Friston, K. (2009) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2009.04.005 · Cited by 1,831
Self-citationGraphical Models, Exponential Families, and Variational Inference
Wainwright, M. J., & Jordan, M. I. (2008) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000001 · Cited by 1,820
Taube, J., Muller, R., & Ranck, J. (1990) · The Journal of Neuroscience · doi:10.1523/jneurosci.10-02-00420.1990 · Cited by 1,809 · cited 1×
“This is the raison d’être of the study of place cells [99], grid cells [100] and head-direction cells [101,102], where the states inferred are physical locations in space.”
Background(2003) · Cited by 1,680
International Encyclopedia of Statistical Science
(2011) · Springer eBooks · doi:10.1007/978-3-642-04898-2 · Cited by 1,653
Constructing Free-Energy Approximations and Generalized Belief Propagation Algorithms
Yedidia, J. S., Freeman, W. T., & Weiss, Y. (2005) · IEEE Transactions on Information Theory · doi:10.1109/tit.2005.850085 · Cited by 1,617
Predictive coding under the free-energy principle
Friston, K., & Kiebel, S. (2009) · Philosophical Transactions of the Royal Society B: Biological Sciences · doi:10.1098/rstb.2008.0300 · Cited by 1,579
Self-citationA free energy principle for the brain
Friston, K., Kilner, J., & Harrison, L. (2006) · Journal of Physiology-Paris · doi:10.1016/j.jphysparis.2006.10.001 · Cited by 1,369
Self-citationActive inference: a process theory
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., & Pezzulo, G. (2017) · Neural Computation · doi:10.1162/neco_a_00912 · Cited by 1,156 · cited 10×
“All other aspects of belief updating (planning as inference, decision-making and learning) were conserved across schemes, following [7,39,57] and their publicly available software implementation (see Software note).”
BackgroundInfluentialSelf-citationInformation Geometry and Its Applications
Amari, S.-I. (2016) · Applied Mathematical Sciences · doi:10.1007/978-4-431-55978-8 · Cited by 1,138
The dynamic brain: from spiking neurons to neural masses and cortical fields
Deco, G., Jirsa, V. K., Robinson, P. A., Breakspear, M., & Friston, K. (2008) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1000092 · Cited by 1,084
Self-citationThe hippocampus as a predictive map
Stachenfeld, K. L., Botvinick, M. M., & Gershman, S. J. (2017) · Nature Neuroscience · doi:10.1038/nn.4650 · Cited by 1,000
An Introduction to factor graphs
Loeliger, H. (2004) · IEEE Signal Processing Magazine · doi:10.1109/msp.2004.1267047 · Cited by 905
(2009) · doi:10.1007/978-3-540-29678-2 · Cited by 900
Optimal control of Markov processes with incomplete state information
Åström, K. J. (1965) · Journal of Mathematical Analysis and Applications · doi:10.1016/0022-247x(65)90154-x · Cited by 854 · cited 2×
“These tasks involve an agent equipped with a generative model (in our case, a partially observed Markov decision process [98]) and a repeated sequence of the following steps [7]:”
BackgroundThe Computational Anatomy of Psychosis
Adams, R. A., Stephan, K. E., Brown, H. R., Frith, C. D., & Friston, K. J. (2013) · Frontiers in Psychiatry · doi:10.3389/fpsyt.2013.00047 · Cited by 850
Self-citationFriston, K. (2013) · Journal of The Royal Society Interface · doi:10.1098/rsif.2013.0475 · Cited by 754
Self-citationEfficient coding of natural sounds
Lewicki, M. S. (2002) · Nature Neuroscience · doi:10.1038/nn831 · Cited by 674
Wiring optimization can relate neuronal structure and function
Chen, B. L., Hall, D. H., & Chklovskii, D. B. (2006) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.0506806103 · Cited by 650
(2005) · Edinburgh Research Explorer (University of Edinburgh) · doi:10.5555/1046920.1088695 · Cited by 650
Redundancy reduction revisited
Barlow, H. (2001) · Network: Computation in Neural Systems · doi:10.1080/net.12.3.241.253 · Cited by 620 · cited 1×
“, Horace Barlow’s principle of minimum redundancy [69]).”
BackgroundNatural image statistics and efficient coding
Olshausen, B. A., & Field, D. J. (1996) · Network: Computation in Neural Systems · doi:10.1088/0954-898x_7_2_014 · Cited by 561
Joyce, J. M. (2011) · International Encyclopedia of Statistical Science · doi:10.1007/978-3-642-04898-2_327 · Cited by 555
Chen, L. L., Lin, L.-H., Green, E. J., Barnes, C. A., & McNaughton, B. L. (1994) · Experimental Brain Research · doi:10.1007/bf00243212 · Cited by 543
Notes on Landauer's principle, reversible computation, and Maxwell's Demon
Bennett, C. H. (2003) · Studies in History and Philosophy of Science Part B: Studies in History and Philosophy of Modern Physics · doi:10.1016/s1355-2198(03)00039-x · Cited by 489 · cited 1×
“Landauer famously observed that a change in information entails heat generation [81,82].”
BackgroundLevy, W. B., & Baxter, R. A. (1996) · Neural Computation · doi:10.1162/neco.1996.8.3.531 · Cited by 481 · cited 1×
“Efficiency of neural processing in biological agents is best seen in the efficient coding hypothesis by Horace Barlow [69–71], which has received much empirical support [65,66,72–77] (see, in particular, [64] for energetic efficiency) and has been a successful driver in computational neural modelling [75,78–80].”
BackgroundDan, Y., Atick, J. J., & Reid, R. C. (1996) · The Journal of Neuroscience · doi:10.1523/jneurosci.16-10-03351.1996 · Cited by 478
Efficient codes and balanced networks
Denève, S., & Machens, C. K. (2016) · Nature Neuroscience · doi:10.1038/nn.4243 · Cited by 471
Measuring Thermodynamic Length
Crooks, G. E. (2007) · Physical Review Letters · doi:10.1103/physrevlett.99.100602 · Cited by 424 · cited 1×
“3 of a path and the energy consumed by travelling along that path [83,84].”
BackgroundThe free energy principle for action and perception: A mathematical review
Buckley, C. L., Kim, C. S., McGregor, S., & Seth, A. K. (2017) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2017.09.004 · Cited by 363 · cited 1×
“This corresponds to performing approximate Bayesian inference (also known as variational Bayes) [2,3,8–11], a standard method in machine learning, or minimising the discrepancy between predictions and sensations [1,12].”
MethodActive Inference, Curiosity and Insight
Friston, K. J., Lin, M., Frith, C. D., Pezzulo, G., Hobson, J. A., & Ondobaka, S. (2017) · Neural Computation · doi:10.1162/neco_a_00999 · Cited by 351 · cited 7×
“A complete specification of the generative model for this rule learning task may be found in [39].”
BackgroundInfluentialSelf-citationNGBoost: Natural Gradient Boosting for Probabilistic PredictionPreprint
Duan, T., Avati, A., Ding, D. Y., Thai, K. K., Basu, S., Ng, A. Y., & Schuler, A. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1910.03225 · Cited by 346
With or without you: predictive coding and Bayesian inference in the brain
Aitchison, L., & Lengyel, M. (2017) · Current Opinion in Neurobiology · doi:10.1016/j.conb.2017.08.010 · Cited by 308
Optimal control as a graphical model inference problem
Kappen, H. J., Gómez, V., & Opper, M. (2012) · Machine Learning · doi:10.1007/s10994-012-5278-7 · Cited by 297
A tutorial on the free-energy framework for modelling perception and learning
Bogacz, R. (2017) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2015.11.003 · Cited by 283 · cited 2×
“The last decade has seen the development of a theory of how the brain might implement active inference consistently in neurobiology [1,7,49,57,58].”
BackgroundMethodAy, N., Jost, J., Lê, H. V., & Schwachhöfer, L. (2017) · Ergebnisse der Mathematik und ihrer Grenzgebiete 34 · doi:10.1007/978-3-319-56478-4 · Cited by 268
Fisher-Rao Metric, Geometry, and Complexity of Neural NetworksPreprint
Liang, T., Poggio, T., Rakhlin, A., & Stokes, J. (2017) · International Conference on Artificial Intelligence and Statistics · doi:10.48550/arxiv.1711.01530 · Cited by 265
Neural masses and fields in dynamic causal modeling
Moran, R., Pinotsis, D. A., & Friston, K. (2013) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2013.00057 · Cited by 255 · cited 1×
“Currently, the sigmoid activation function is the most commonly used function to relate average transmembrane potential to average firing rate in mean-field formulations of neural population dynamics [91,92] and deep neural networks [93,94].”
MethodSelf-citationComputational mechanisms of curiosity and goal-directed exploration
Schwartenbeck, P., Passecker, J., Hauser, T. U., FitzGerald, T. H., Kronbichler, M., & Friston, K. J. (2019) · eLife · doi:10.7554/elife.41703 · Cited by 254
Self-citationUncertainty, epistemics and active inference
Parr, T., & Friston, K. J. (2017) · Journal of The Royal Society Interface · doi:10.1098/rsif.2017.0376 · Cited by 252
Self-citationActive Inference: Demystified and Compared
Sajid, N., Ball, P. J., Parr, T., & Friston, K. J. (2021) · Neural Computation · doi:10.1162/neco_a_01357 · Cited by 252 · cited 1×
“Active inference agents show competitive or state-of-the-art performance in a wide variety of simulated environments [34,46,47,56].”
BackgroundSelf-citationActive inference on discrete state-spaces: A synthesis
Da Costa, L., Parr, T., Sajid, N., Veselic, S., Neacsu, V., & Friston, K. (2020) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2020.102447 · Cited by 226
Self-citationA free energy principle for a particular physics
Friston, K. (2019) · arXiv · doi:10.48550/arxiv.1906.10184 · Cited by 222 · cited 1×
“2 certain systems at steady-state as having the appearance of sentience [5,6].”
BackgroundSelf-citationThe dopaminergic midbrain encodes the expected certainty about desired outcomes
Schwartenbeck, P., FitzGerald, T. H. B., Mathys, C., Dolan, R., & Friston, K. (2015) · Cerebral Cortex · doi:10.1093/cercor/bhu159 · Cited by 219 · cited 1×
“A number of predictions of this theory have been empirically validated, including the role of dopamine in decisionmaking [59,60] and free energy minimisation in exploration and choice behaviour [61,62].”
BackgroundSelf-citationInformation and efficiency in the nervous system-a synthesis.
Sengupta, B., Stemmler, M. B., & Friston, K. J. (2013) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1003157 · Cited by 212 · cited 1×
“This means that to infer the states of their environment, biological agents reduce the discrepancy between their predictions of the environment and their observations, or maximise the mutual information between them [63].”
BackgroundSelf-citationExploration, novelty, surprise, and free energy minimization
Schwartenbeck, P., FitzGerald, T., Dolan, R. J., & Friston, K. (2013) · Frontiers in Psychology · doi:10.3389/fpsyg.2013.00710 · Cited by 207 · cited 1×
“A small number of predictions of this theory have been empirically validated, including the role of dopamine in decision-making [59,60] and free energy minimisation in exploration and choice behaviour [61,62].”
BackgroundSelf-citationThe Anatomy of Inference: Generative Models and Brain Structure
Parr, T., & Friston, K. J. (2018) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2018.00090 · Cited by 204 · cited 1×
“The last decade has developed a theory of how the brain might be implementing active inference consistently with neurology [1,7,49,57,58].”
BackgroundSelf-citationPlanning and navigation as active inference
Kaplan, R., & Friston, K. J. (2018) · Biological Cybernetics · doi:10.1007/s00422-018-0753-2 · Cited by 203
Self-citationSmith, R., Kuplicki, R., Feinstein, J., Forthman, K. L., Stewart, J. L., Paulus, M. P., Tulsa 1000 investigators, & Khalsa, S. S. (2020) · PLOS Computational Biology · doi:10.1371/journal.pcbi.1008484 · Cited by 196
Friston, K. J., Costa, L. D., Hafner, D., Hesp, C., & Parr, T. (2021) · Neural Computation · doi:10.1162/neco_a_01351 · Cited by 183
Self-citationPopulation dynamics: Variance and the sigmoid activation function
Marreiros, A. C., Daunizeau, J., Kiebel, S. J., & Friston, K. J. (2008) · NeuroImage · doi:10.1016/j.neuroimage.2008.04.239 · Cited by 176 · cited 2×
“Marreiros and colleagues showed that the sigmoid activation function can be more plausibly motivated by considering the variance of neuronal potentials within a population [88], which is generally modelled by a Gaussian distribution under the Laplace assumption in mean-field treatments of neural population dynamics [8…”
BackgroundSelf-citationOn Variational Message Passing on Factor Graphs
Dauwels, J. (2007) · 2007 IEEE International Symposium on Information Theory · doi:10.1109/isit.2007.4557602 · Cited by 174
Scene Construction, Visual Foraging, and Active Inference
Mirza, M. B., Adams, R. A., Mathys, C. D., & Friston, K. J. (2016) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2016.00056 · Cited by 165
Self-citationStochastic Thermodynamic Interpretation of Information Geometry
Ito, S. (2018) · Physical Review Letters · doi:10.1103/physrevlett.121.030605 · Cited by 162 · cited 1×
“3 of a path and the energy consumed by travelling along that path [83,84].”
BackgroundCell‐attached measurements of the firing threshold of rat hippocampal neurones
Fricker, D., Verheugen, J. A. H., & Miles, R. (1999) · The Journal of Physiology · doi:10.1111/j.1469-7793.1999.0791s.x · Cited by 160 · cited 1×
“However, the dispersion of neuronal thresholds is, quantitatively, much less important than the variance of neuronal membrane potential within populations [87].”
BackgroundEfficient coding in heterogeneous neuronal populations
Chelaru, M. I., & Dragoi, V. (2008) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.0807744105 · Cited by 158
Amari, S., & Douglas, S. C. (2002) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.1998.675489 · Cited by 157
Polytomous logistic regression
Engel, J. (1988) · Statistica Neerlandica · doi:10.1111/j.1467-9574.1988.tb01238.x · Cited by 144
Markov blankets, information geometry and stochastic thermodynamics
Parr, T., Da Costa, L., & Friston, K. (2020) · Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences · doi:10.1098/rsta.2019.0159 · Cited by 141 · cited 1×
“Active inference is a normative framework for explaining behaviour under the free energy principle, a theory of self-organisation originating in neuroscience [1–4] that characterises certain systems at steady-state as having the appearance of sentience [5,6].”
BackgroundSelf-citationReinforcement Learning through Active InferencePreprint
Tschantz, A., Millidge, B., Seth, A. K., & Buckley, C. L. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.12636 · Cited by 140 · cited 1×
“Active inference agents show competitive or state-of-the-art performance in a wide variety of simulated environments [34,46,47,56].”
BackgroundAn Active Inference Approach to Modeling Structure Learning: Concept Learning as an Example Case
Smith, R. S., Schwartenbeck, P., Parr, T., & Friston, K. J. (2020) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2020.00041 · Cited by 132
Self-citationNeuronal message passing using mean-field, bethe, and marginal approximations
Parr, T., Markovic, D., Kiebel, S. J., & Friston, K. J. (2019) · Scientific Reports · doi:10.1038/s41598-018-38246-3 · Cited by 128
Self-citationLearning action-oriented models through active inference
Tschantz, A., Seth, A. K., & Buckley, C. L. (2020) · PLOS Computational Biology · doi:10.1371/journal.pcbi.1007805 · Cited by 121
Deep active inference as variational policy gradients
Millidge, B. (2020) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2020.102348 · Cited by 115
Dopamine, reward learning, and active inference
FitzGerald, T. H. B., Dolan, R. J., & Friston, K. (2015) · Frontiers in Computational Neuroscience · doi:10.3389/fncom.2015.00136 · Cited by 112 · cited 1×
“A number of predictions of this theory have been empirically validated, including the role of dopamine in decisionmaking [59,60] and free energy minimisation in exploration and choice behaviour [61,62].”
BackgroundSelf-citationFirst principles in the life sciences: the free-energy principle, organicism, and mechanism
Colombo, M., & Wright, C. (2021) · Synthese · doi:10.1007/s11229-018-01932-w · Cited by 101
Ueltzhöffer, K. (2018) · Biological Cybernetics · doi:10.1007/s00422-018-0785-7 · Cited by 98
On the Relationship Between Active Inference and Control as Inference
Millidge, B., Tschantz, A., Seth, A. K., & Buckley, C. L. (2020) · Communications in computer and information science · doi:10.1007/978-3-030-64919-7_1 · Cited by 97
Evidence for surprise minimization over value maximization in choice behavior
Schwartenbeck, P., FitzGerald, T. H. B., Mathys, C., Dolan, R., Kronbichler, M., & Friston, K. (2015) · Scientific Reports · doi:10.1038/srep16575 · Cited by 96 · cited 1×
“A number of predictions of this theory have been empirically validated, including the role of dopamine in decisionmaking [59,60] and free energy minimisation in exploration and choice behaviour [61,62].”
BackgroundSelf-citationVision and the Coding of Natural Images
Olshausen, B. A., & Field, D. (2000) · American Scientist · doi:10.1511/2000.3.238 · Cited by 95
The Wiring Economy Principle: Connectivity Determines Anatomy in the Human Brain
Raj, A., & Chen, Y.-H. (2011) · PLoS ONE · doi:10.1371/journal.pone.0014832 · Cited by 90
Population dynamics under the Laplace assumption
Marreiros, A. C., Kiebel, S. J., Jean Daunizeau, A., Harrison, L. M., & Friston, K. J. (2008) · NeuroImage · doi:10.1016/j.neuroimage.2008.10.008 · Cited by 89 · cited 1×
“Marreiros and colleagues showed that the sigmoid activation function can be more plausibly motivated by considering the variance of neuronal potentials within a population [88], which is generally modelled by a Gaussian distribution under the Laplace assumption in mean-field treatments of neural population dynamics [8…”
BackgroundSelf-citationEfficient coding of natural images with a population of noisy Linear-Nonlinear neurons.
Karklin, Y., & Simoncelli, E. P. (2011) · PubMed · Cited by 88
Coding of position by simultaneously recorded sensory neurones in the cat dorsal root ganglion
Stein, R. B., Weber, D. J., Aoyagi, Y., Procházka, A., Wagenaar, J., Shoham, S., & Normann, R. A. (2004) · The Journal of Physiology · doi:10.1113/jphysiol.2004.068668 · Cited by 87
Human visual exploration reduces uncertainty about the sensed world
Mirza, M. B., Adams, R. A., Mathys, C. D., & Friston, K. J. (2018) · PLoS ONE · doi:10.1371/journal.pone.0190429 · Cited by 87
Self-citationSmith, R. S., Schwartenbeck, P., Stewart, J. L., Kuplicki, R., Ekhtiari, H., T Investigators, & Paulus, M. P. (2020) · Drug and Alcohol Dependence · doi:10.1016/j.drugalcdep.2020.108208 · Cited by 84
Adaptive Robot Body Learning and Estimation Through Predictive Coding
Lanillos, P., & Cheng, G. (2018) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2018.8593684 · Cited by 73 · cited 1×
“These include planning and navigation [25– 31], exploration [32–36], learning of self and others [37,38], concept learning [39–41], playing games [42–44], adapting to changing environments [45–47], active vision [48–51] and psychiatric illness [42,52–55].”
BackgroundNatural gradient algorithm for blind separation of overdetermined mixture with additive noise
Zhang, L.-Q., Cichocki, A. S., & Амари, Ш.-И. (1999) · IEEE Signal Processing Letters · doi:10.1109/97.796292 · Cited by 72
Active Inference in OpenAI Gym: A Paradigm for Computational Investigations Into Psychiatric Illness
Cullen, M., Davey, B., Friston, K. J., & Moran, R. (2018) · Biological Psychiatry Cognitive Neuroscience and Neuroimaging · doi:10.1016/j.bpsc.2018.06.010 · Cited by 70
Self-citationAn Active Inference view of cognitive control
Pezzulo, G. (2012) · Frontiers in Psychology · doi:10.3389/fpsyg.2012.00478 · Cited by 68
Exact natural gradient in deep linear networks and its application to the nonlinear case
Bernacchia, A., Lengyel, M., & Hennequin, G. (2018) · neural information processing systems · Cited by 66
Deep active inference agents using Monte-Carlo methods
Fountas, Z., Sajid, N., Mediano, P. A. M., & Friston, K. (2020) · arXiv · doi:10.48550/arxiv.2006.04176 · Cited by 64
Self-citationWeber, D. J., Stein, R. B., Everaert, D. G., & Procházka, A. (2006) · IEEE Transactions on Neural Systems and Rehabilitation Engineering · doi:10.1109/tnsre.2006.875575 · Cited by 58
End-to-End Pixel-Based Deep Active Inference for Body Perception and Action
Sancaktar, C., van Gerven, M. A. J., & Lanillos, P. (2020) · 2020 Joint IEEE 10th International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob) · doi:10.1109/icdl-epirob48136.2020.9278105 · Cited by 58
Active inference and the anatomy of oculomotion
Parr, T., & Friston, K. J. (2018) · Neuropsychologia · doi:10.1016/j.neuropsychologia.2018.01.041 · Cited by 57
Self-citationEfficient Olfactory Coding in the Pheromone Receptor Neuron of a Moth
Kostal, L., Lánský, P., & Rospars, J.-P. (2008) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1000053 · Cited by 56
An empirical evaluation of active inference in multi-armed bandits
Marković, D., Stojić, H., Schwöbel, S., & Kiebel, S. J. (2021) · Neural Networks · doi:10.1016/j.neunet.2021.08.018 · Cited by 51 · cited 1×
“Active inference agents show competitive or state-of-the-art performance in a wide variety of simulated environments [34,46,47,56].”
BackgroundActive Inference, Belief Propagation, and the Bethe Approximation
Schwöbel, S., Kiebel, S. J., & Marković, D. (2018) · Neural Computation · doi:10.1162/neco_a_01108 · Cited by 47 · cited 1×
“Once the appropriate generative model is found, one would need to compare the evidence for a few possible practical implementations of active inference, which come from various possible approximations to the free energy [27,124,125], each of which yields different belief updates and simulated electrophysiological resp…”
BackgroundLearning Perception and Planning With Deep Active Inference
Çatal, O., Verbelen, T., Nauta, J., Boom, C. D., & Dhoedt, B. (2020) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp40776.2020.9054364 · Cited by 42
Matsumoto, T., & Tani, J. (2020) · Entropy · doi:10.3390/e22050564 · Cited by 42
Smith, R. S., Kuplicki, R., Feinstein, J. S., et al. (2020) · medRxiv · doi:10.1101/2020.06.03.20121343 · Cited by 38
Robot self/other distinction: active inference meets neural networks learning in a mirror
Lanillos, P., Pages, J., & Cheng, G. (2020) · arXiv · doi:10.48550/arxiv.2004.05473 · Cited by 35 · cited 1×
“These include planning and navigation [25– 31], exploration [32–36], learning of self and others [37,38], concept learning [39–41], playing games [42–44], adapting to changing environments [45–47], active vision [48–51] and psychiatric illness [42,52–55].”
BackgroundReward maximization through discrete active inference
Da Costa, L., Sajid, N., Parr, T., Friston, K., & Smith, R. (2023) · Neural Computation · doi:10.1162/neco_a_01574 · Cited by 35
Self-citationState-space decoding of primary afferent neuron firing rates
Wagenaar, J., Ventura, V., & Weber, D. J. (2011) · Journal of Neural Engineering · doi:10.1088/1741-2560/8/1/016002 · Cited by 30
Marković, D., Goschke, T., & Kiebel, S. J. (2020) · Cognitive Affective & Behavioral Neuroscience · doi:10.3758/s13415-020-00837-x · Cited by 29
Olshausen, B. A., & O’Connor, K. N. (2002) · Nature Neuroscience · doi:10.1038/nn0402-292 · Cited by 25
Smith, R. S., Schwartenbeck, P., Stewart, J. L., et al. (2020) · doi:10.31234/osf.io/a794k · Cited by 16
Deep Active Inference for Autonomous Robot NavigationPreprint
Çatal, O., Wauthier, S. T., Verbelen, T., Boom, C. D., & Dhoedt, B. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.03220 · Cited by 16
Natural gradient‐projection algorithm for distribution control
Zhang, Z., Sun, H., & Zhong, F. (2008) · Optimal Control Applications and Methods · doi:10.1002/oca.874 · Cited by 15
Iterative scaling and coordinate descent methods for maximum entropy
Huang, F.-L., Hsieh, C., Chang, K., & Lin, C. (2009) · Joint Conference of the Annual Meeting of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (ACL-IJCNLP) · doi:10.3115/1667583.1667671 · Cited by 15
Marković, D., Reiter, A. M. F., & Kiebel, S. J. (2019) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1006707 · Cited by 14
Sleep: Model Reduction in Deep Active Inference
Wauthier, S. T., Çatal, O., Boom, C. D., Verbelen, T., & Dhoedt, B. (2020) · Communications in computer and information science · doi:10.1007/978-3-030-64919-7_9 · Cited by 11
The computational neurology of active vision
Parr, T. (2019) · UCL Discovery (University College London) · Cited by 11
Self-citationSmith, R. S., Kirlić, N., Stewart, J. L., Touthang, J., Kuplicki, R., Khalsa, S. S., Paulus, M. P., T Investigators, & Aupperle, R. L. (2020) · PsyArXiv (OSF Preprints) · doi:10.31234/osf.io/t2dhn · Cited by 9
Marković, D., Goschke, T., & Kiebel, S. J. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/847566 · Cited by 7
Active Inference or Control as Inference? A Unifying ViewPreprint
Watson, J., Imohiosen, A., & Peters, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.00262 · Cited by 4
The projection adaptive natural gradient online algorithm for SVM
Sun, Z., & Shi, B. (2010) · Chinese Control Conference (CCC) · Cited by 3
(2008) · Springer eBooks · doi:10.1007/978-3-540-29678-2_2901
End-to-End Pixel-Based Deep Active Inference for Body Perception and Action
Sancaktar, C., van Gerven, M., & Lanillos, P. (2020) · arXiv · doi:10.48550/arxiv.2001.05847
On the Relationship Between Active Inference and Control as Inference
Millidge, B., Tschantz, A., Seth, A. K., & Buckley, C. L. (2020) · arXiv · doi:10.48550/arxiv.2006.12964
An elementary introduction to information geometry
Nielsen, F. (2018) · arXiv · doi:10.48550/arxiv.1808.08271
Research works printed but not linked: 4
- Da Costa, L., Sajid, N., Parr, T., Friston, K., and Smith, R. (2021, January 31). The Relationship between Dynamic Programming and Active Inference: The Discrete, Finite-Horizon Case. arXiv:200908111 [cs, math, q-bio]. Available online: http://arxiv.org/abs/2009.08111. · arXiv
- Barlow, H.B. (2021, April 09). Possible Principles Underlying the Transformations of Sensory Messages; The MIT Press: 1961. Available online: https://www.universitypressscholarship.com/view/10.7551/mitpress/9780262518420.001.0001/upso-9780262518420-chapter-13. · doi:10.7551/mitpress/9780262518420.001.0001/upso-9780262518420-chapter-13
- Bishop, C.M. (2006). Pattern Recognition and Machine Learning, Springer.
- Stone, J.V. (2019). Artificial Intelligence Engines: A Tutorial Introduction to the Mathematics of Deep Learning, Sebtel Press.
Cited by
A step-by-step tutorial on active inference and its application to empirical data
Smith, R., Friston, K. J., & Whyte, C. J. (2022) · Journal of Mathematical Psychology · doi:10.1016/j.jmp.2021.102632 · Cited by 246
Self-citationThe free energy principle made simpler but not too simple
Friston, K., Da Costa, L., Sajid, N., Heins, C., Ueltzhöffer, K., Pavliotis, G. A., & Parr, T. (2023) · Physics Reports · doi:10.1016/j.physrep.2023.07.001 · Cited by 163
“Indeed, the Fisher information metric tensor, which measures changes in the Kullback–Leibler (KL) divergence resulting from infinitesimal changes in the internal mode, is a Riemannian metric that yields an information distance [42, Appendix B].”
Self-citationComputational psychiatry: from synapses to sentience
Friston, K. J. (2022) · Molecular Psychiatry · doi:10.1038/s41380-022-01743-z · Cited by 189
Self-citationHow particular is the physics of the free energy principle?
Aguilera, M., Millidge, B., Tschantz, A., & Buckley, C. L. (2021) · Physics of Life Reviews · doi:10.1016/j.plrev.2021.11.001 · Cited by 103
Predictive coding is a consequence of energy efficiency in recurrent neural networks
Ali, A. G., Ahmad, N., de Groot, E., van Gerven, M. A. J., & Kietzmann, T. C. (2022) · Patterns · doi:10.1016/j.patter.2022.100639 · Cited by 104
Metadata
- Authors
- Da Costa, L., Parr, T., Sengupta, B., & Friston, K.
- Affiliations
- Department of Mathematics, Imperial College London, London SW7 2AZ, UK
- Wellcome Centre for Human Neuroimaging (GB)
- Venue
- Entropy
- Publisher
- MDPI AG
- Published
- 2021-04-12
- Volume
- 23
- Issue
- 4
- Pages
- 454
- ISSN
- 1099-4300 (ISSN-L)
- PMC
- PMC8069154
- OpenAlex
- W3126619093
- Cited by
- 38
- References
- 110
- Citations by year
- 2026: 2 · 2025: 7 · 2024: 5 · 2023: 9 · 2022: 7 · 2021: 8
- FWCI
- 3.81
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Entropy (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Funders
- Fonds National de la Recherche Luxembourg (13568875)
- Wellcome Trust (088130/Z/09/Z)
- Rosetrees Trust (173346)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
- Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
- Life Sciences › Neuroscience › Cognitive Neuroscience › Cognitive Science and Education Research
- Keywords
- active inference, free energy principle, neural dynamics, prediction error, natural gradient descent, information geometry, information length, metabolic cost, belief updating, generative models, electrophysiological responses, neural representations, state estimation
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
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.