Academia.eduAcademia.edu

Outline

Title
Abstract
References
All Topics
Computer Science

Flexible intentions: An Active Inference theory

Profile image of Ivilin P StoianovIvilin P Stoianov

2023, Frontiers in Computational Neuroscience

https://doi.org/10.3389/FNCOM.2023.1128694
visibility

…

description

27 pages

link

1 file

Abstract

We present a normative computational theory of how the brain may support visually-guided goal-directed actions in dynamically changing environments. It extends the Active Inference theory of cortical processing according to which the brain maintains beliefs over the environmental state, and motor control signals try to fulfill the corresponding sensory predictions. We propose that the neural circuitry in the Posterior Parietal Cortex (PPC) compute flexible intentions-or motor plans from a belief over targets-to dynamically generate goal-directed actions, and we develop a computational formalization of this process. A proof-of-concept agent embodying visual and proprioceptive sensors and an actuated upper limb was tested on target-reaching tasks. The agent behaved correctly under various conditions, including static and dynamic targets, di erent sensory feedbacks, sensory precisions, intention gains, and movement policies; limit conditions were individuated, too. Active Inference driven by dynamic and flexible intentions can thus support goal-directed behavior in constantly changing environments, and the PPC might putatively host its core intention mechanism. More broadly, the study provides a normative computational basis for research on goal-directed behavior in end-to-end settings and further advances mechanistic theories of active biological systems.

Related papers

Predictions not commands: active inference in the motor system
Rick Adams, S. Shipp

Brain Structure and Function, 2013

The descending projections from motor cortex share many features with top-down or backward connections in visual cortex; for example, corticospinal projections originate in infragranular layers, are highly divergent and (along with descending cortico-cortical projections) target cells expressing NMDA receptors. This is somewhat paradoxical because backward modulatory characteristics would not be expected of driving motor command signals. We resolve this apparent paradox using a functional characterisation of the motor system based on Helmholtz's ideas about perception; namely, that perception is inference on the causes of visual sensations. We explain behaviour in terms of inference on the causes of proprioceptive sensations. This explanation appeals to active inference, in which higher cortical levels send descending proprioceptive predictions, rather than motor commands. This process mirrors perceptual inference in sensory cortex, where descending connections convey predictions, while ascending connections convey prediction errors. The anatomical substrate of this recurrent message passing is a hierarchical system consisting of functionally asymmetric driving (ascending) and modulatory (descending) connections: an arrangement that we show is almost exactly recapitulated in the motor system, in terms of its laminar, topographic and physiological characteristics. This perspective casts classical motor reflexes as minimising prediction errors and may provide a principled explanation for why motor cortex is agranular.

View PDFchevron_right
Action understanding and active inference
Jérémie Mattout

Biological Cybernetics, 2011

We have suggested that the mirror-neuron system might be usefully understood as implementing Bayes-optimal perception of actions emitted by oneself or others. To substantiate this claim, we present neuronal simulations that show the same representations can prescribe motor behavior and encode motor intentions during action–observation. These simulations are based on the free-energy formulation of active inference, which is formally related to predictive coding. In this scheme, (generalised) states of the world are represented as trajectories. When these states include motor trajectories they implicitly entail intentions (future motor states). Optimizing the representation of these intentions enables predictive coding in a prospective sense. Crucially, the same generative models used to make predictions can be deployed to predict the actions of self or others by simply changing the bias or precision (i.e. attention) afforded to proprioceptive signals. We illustrate these points using simulations of handwriting to illustrate neuronally plausible generation and recognition of itinerant (wandering) motor trajectories. We then use the same simulations to produce synthetic electrophysiological responses to violations of intentional expectations. Our results affirm that a Bayes-optimal approach provides a principled framework, which accommodates current thinking about the mirror-neuron system. Furthermore, it endorses the general formulation of action as active inference.

View PDFchevron_right
Planning Movements in Visual and Physical Space in Monkey Posterior Parietal Cortex
Alexander Gail

Cerebral cortex (New York, N.Y. : 1991), 2015

Neurons in the posterior parietal cortex respond selectively for spatial parameters of planned goal-directed movements. Yet, it is still unclear which aspects of the movement the neurons encode: the spatial parameters of the upcoming physical movement (physical goal), or the upcoming visual limb movement (visual goal). To test this, we recorded neuronal activity from the parietal reach region while monkeys planned reaches under either normal or prism-reversed viewing conditions. We found predominant encoding of physical goals while fewer neurons were selective for visual goals during planning. In contrast, local field potentials recorded in the same brain region exhibited predominant visual goal encoding, similar to previous imaging data from humans. The visual goal encoding in individual neurons was neither related to immediate visual input nor to visual memory, but to the future visual movement. Our finding suggests that action planning in parietal cortex is not exclusively a prec...

View PDFchevron_right
Parietal maps of visual signals for bodily action planning
Alessia Sepe

Brain structure & function, 2021

The posterior parietal cortex (PPC) has long been understood as a high-level integrative station for computing motor commands for the body based on sensory (i.e., mostly tactile and visual) input from the outside world. In the last decade, accumulating evidence has shown that the parietal areas not only extract the pragmatic features of manipulable objects, but also subserve sensorimotor processing of others' actions. A paradigmatic case is that of the anterior intraparietal area (AIP), which encodes the identity of observed manipulative actions that afford potential motor actions the observer could perform in response to them. On these bases, we propose an AIP manipulative action-based template of the general planning functions of the PPC and review existing evidence supporting the extension of this model to other PPC regions and to a wider set of actions: defensive and locomotor actions. In our model, a hallmark of PPC functioning is the processing of information about the phy...

View PDFchevron_right
The posterior parietal cortex: Sensorimotor interface for the planning and online control of visually guided movements
Christopher Buneo

Neuropsychologia, 2006

We present a view of the posterior parietal cortex (PPC) as a sensorimotor interface for visually guided movements. Special attention is given to the role of the PPC in arm movement planning, where representations of target position and current hand position in an eye-centered frame of reference appear to be mapped directly to a representation of motor error in a hand-centered frame of reference. This mapping is direct in the sense that it does not require target position to be transformed into intermediate reference frames in order to derive a motor error signal in hand-centered coordinates. Despite being direct, this transformation appears to manifest in the PPC as a gradual change in the functional properties of cells along the ventro-dorsal axis of the superior parietal lobule (SPL), i.e. from deep in the sulcus to the cortical surface. Possible roles for the PPC in context dependent coordinate transformations, formation of intrinsic movement representations, and in online control of visually guided arm movements are also discussed. Overall these studies point to the emerging view that, for arm movements, the PPC plays a role not only in the inverse transformations required to convert sensory information into motor commands but also in 'forward' transformations as well, i.e. in integrating sensory input with previous and ongoing motor commands to maintain a continuous estimate of arm state that can be used to update present and future movement plans. Critically, this state estimate appears to be encoded in an eye-centered frame of reference.

View PDFchevron_right
Neural Correlates of Reaching Decisions in Dorsal Premotor Cortex: Specification of Multiple Direction Choices and Final Selection of Action
Paul Cisek, John Kalaska

Neuron, 2005

options the one that will be released into overt execution.

View PDFchevron_right
dorsal parietofrontal cortex Effector selection precedes reach planning in the
Eric Yttri

2014

imental evidence and computational modeling suggest that target selection for reaching is associated with the parallel encoding of multiple movement plans in the dorsomedial posterior parietal cortex (dmPPC) and the caudal part of the dorsal premotor cortex (PMdc). We tested the hypothesis that a similar mechanism also accounts for arm selection for unimanual reaching, with simultaneous and separate motor goal representations for the left and right arms existing in the right and left parietofrontal cortex, respectively. We recorded simultaneous electroencephalograms and functional MRI and studied a condition in which subjects had to select the appropriate arm for reaching based on the color of an appearing visuospatial target, contrasting it to a condition in which they had full knowledge of the arm to be used before target onset. We showed that irrespective of whether subjects had to select the arm or not, activity in dmPPC and PMdc was only observed contralateral to the reaching arm after target onset. Furthermore, the latency of activation in these regions was significantly delayed when arm selection had to be achieved during movement planning. Together, these results demonstrate that effector selection is not achieved through the simultaneous specification of motor goals tied to the two arms in bilateral parietofrontal cortex, but suggest that a motor goal is formed in these regions only after an arm is selected for action.

View PDFchevron_right
Conditional visuomotor behavior in the Parietal Reach Region and Dorsal Premotor Cortex
Christian Klaes

Planning goal-directed movements requires the combination of visuospatial with abstract contextual information. Our sensory environment constrains possible movements to a certain extent. However, contextual information guides proper choice of action in a given situation and allows flexible mapping of sensory instruction cues onto different motor actions. We used anti-reach tasks to test the hypothesis that spatial motor-goal representations in cortical sensorimotor areas are gain modulated by the behavioral context to achieve flexible remapping of spatial cue information onto arbitrary motor goals. We found that gain modulation of neuronal reach goal representations is commonly induced by the behavioral context in individual neurons of both, the parietal reach region (PRR) and the dorsal premotor cortex (PMd). In addition, PRR showed stronger directional selectivity during the planning of a reach toward a directly cued goal (pro-reach) compared with an inferred target (anti-reach). PMd, however, showed stronger overall activity during reaches toward inferred targets compared with directly cued targets. Based on our experimental evidence, we suggest that gain modulation is the computational mechanism underlying the integration of spatial and contextual information for flexible, rule-driven stimulus-response mapping, and thereby forms an important basis of goal-directed behavior. Complementary contextual effects in PRR versus PMd are consistent with the idea that posterior parietal cortex preferentially represents sensory-driven, "automatic" motor goals, whereas frontal sensorimotor areas are stronger engaged in the representation of rule-based, "inferred" motor goals.

View PDFchevron_right
Intentions and motor representations: The interface challenge
Elisabeth Pacherie, Myrto Mylopoulos

A full account of purposive action must appeal not only to propositional attitude states like beliefs, desires, and intentions, but also to motor representations, i.e., non-propositional states that are thought to represent, among other things, action outcomes as well as detailed kinematic features of bodily movements. This raises the puzzle of how it is that these two distinct types of state successfully coordinate. We examine this so-called "Interface Problem". First, we clarify and expand on the nature and role of motor representations in explaining intentional action. Next, we characterize the respective functions of intentions and motor representations, the differences in representational format and content that these imply, and the interface challenge these differences in turn raise. We then evaluate Butterfill & Sinigaglia’s (2014) recent answer to this interface challenge, according to which intentions refer to action outcomes by way of demonstrative deference to motor representations. We present some worries for this proposal, arguing that, among other things, it implicitly presupposes a solution to the problem, and so cannot help to resolve it. Finally, we suggest that we may make some progress on this puzzle by positing a “content-preserving causal process” taking place between intentions and motor representations, and we offer a proposal for how this might work.

View PDFchevron_right
Ventral Premotor and Inferior Parietal Cortices Make Distinct Contribution to Action Organization and Intention Understanding
Stefano Rozzi, Luca Bonini

Cerebral Cortex, 2010

It is well known that ventral premotor area F5 codes the goal of executed and observed motor acts. This area is anatomically connected with part of the inferior parietal cortex (area PFG), which has been recently shown to play a role in action organization and intention understanding. The aims of the present study were 1) to assess whether the discharge of F5 motor neurons and mirror neurons (MNs) codes action goals and 2) to clarify the relative contribution of F5 and PFG in action organization and intention understanding. To this purpose, we first recorded from F5 motor neurons and MNs of 2 monkeys while performing a motor task constituted by 2 actions (''grasp-to-eat'' and ''grasp-to-place'') or observing the same task done by an experimenter. Results showed that some F5 neurons code grasping according to the goal of the action in which it is embedded. Subsequently, we recorded from PFG motor neurons and MNs of the same monkeys, using the same tasks. The comparison between the neuronal properties of F5 and PFG motor neurons suggests that PFG plays a major role in organizing natural actions. Furthermore, the similarities between MNs properties of the 2 areas indicate that they constitute a functional circuit underlying others' intention understanding.

View PDFchevron_right

References (88)

  1. Adams, R. A., Aponte, E., Marshall, L., and Friston, K. J. (2015). Active inference and oculomotor pursuit: the dynamic causal modelling of eye movements. J. Neurosci. Methods 242, 1-14. doi: 10.1016/j.jneumeth.2015.
  2. Adams, R. A., Shipp, S., and Friston, K. J. (2013). Predictions not commands: active inference in the motor system. Brain Struct. Funct. 218, 611-643. doi: 10.1007/s00429-012-0475-5
  3. Adams, R. A., Vincent, P., Benrimoh, D., Friston, K. J., and Parr, T. (2021). Everything is connected: Inference and attractors in delusions. Schizophrenia Res. 245, 5-22. doi: 10.1016/j.schres.2021.07.032
  4. Andersen, R. A. (1995). Encoding of intention and spatial location in the posterior parietal cortex. Cereb. Cortex 5, 457-469. doi: 10.1093/cercor/5.5.457
  5. Baioumy, M., Duckworth, P., Lacerda, B., and Hawes, N. (2020). Active inference for integrated state-estimation, control, and learning. arXiv. doi: 10.1109/ICRA48506.2021.9562009
  6. Baldauf, D., Cui, H., and Andersen, R. A. (2008). The posterior parietal cortex encodes in parallel both goals for double-reach sequences. J. Neurosci. 28, 10081-10089. doi: 10.1523/JNEUROSCI.3423-08.2008
  7. Baltieri, M., and Buckley, C. L. (2019). PID control as a process of active inference with linear generative models. Entropy 21, 257. doi: 10.3390/e210
  8. Bastos, A. M., Usrey, W. M., Adams, R. A., Mangun, G. R., Fries, P., and Friston, K. J. (2012). Canonical microcircuits for predictive coding. Neuron 76, 695-711. doi: 10.1016/j.neuron.2012.10.038
  9. Bishop, C. M. (2006). Pattern Recognition and Machine Learning. New York, NY: Springer.
  10. Bogacz, R. (2017). A tutorial on the free-energy framework for modelling perception and learning. J. Math. Psychol. 76, 198-211. doi: 10.1016/j.jmp.2015.11.003
  11. Breveglieri, R., Galletti, C., Dal Bò, G., Hadjidimitrakis, K., and Fattori, P. (2014). Multiple aspects of neural activity during reaching preparation in the medial posterior parietal area V6A. J. Cogn. Neurosci. 26, 879-895. doi: 10.1162/jocn_a_00510
  12. Buckley, C. L., Kim, C. S., McGregor, S., and Seth, A. K. (2017). The free energy principle for action and perception: a mathematical review. J. Math. Psychol. 81, 55-79. doi: 10.1016/j.jmp.2017.09.004
  13. Cisek, P., and Kalaska, J. F. (2010). Neural mechanisms for interacting with a world full of action choices. Annu. Rev. Neurosci. 33, 269-298. doi: 10.1146/annurev.neuro.051508.135409
  14. Cohen, Y. E., and Andersen, R. A. (2002). A common reference frame for movement plans in the posterior parietal cortex. Nat. Rev. Neurosci. 3, 553-562. doi: 10.1038/nrn873
  15. Corbetta, M., and Shulman, G. L. (2002). Control of goal-directed and stimulus- driven attention in the brain. Nat. Rev. Neurosci. 3, 201-215. doi: 10.1038/nrn755
  16. Desmurget, M., Epstein, C. M., Turner, R. S., Prablanc, C., Alexander, G. E., and Grafton, S. T. (1999). PPC and visually directing reaching to targets. Nature Ne 2, 563-567. doi: 10.1038/9219
  17. Doya, K. (2007). Bayesian Brain: Probabilistic Approaches to Neural Coding. Cambridge, MA: The MIT Press.
  18. Erlhagen, W., and Schöner, G. (2002). Dynamic field theory of movement preparation. Psychol. Rev. 109, 545-572. doi: 10.1037/0033-295X.109.3.545
  19. Fattori, P., Breveglieri, R., Bosco, A., Gamberini, M., and Galletti, C. (2017). Vision for prehension in the medial parietal cortex. Cereb. Cortex 27, 1149-1163. doi: 10.1093/cercor/bhv302
  20. Filippini, M., Breveglieri, R., Ali Akhras, M., Bosco, A., Chinellato, E., and Fattori, P. (2017). Decoding information for grasping from the macaque dorsomedial visual stream. J. Neurosci. 37, 4311-4322. doi: 10.1523/JNEUROSCI.3077-16.2017
  21. Filippini, M., Breveglieri, R., Hadjidimitrakis, K., Bosco, A., and Fattori, P. (2018). Prediction of reach goals in depth and direction from the parietal cortex. Cell Rep. 23, 725-732. doi: 10.1016/j.celrep.2018.03.090
  22. FitzGerald, T. H., Moran, R. J., Friston, K. J., and Dolan, R. J. (2015). Precision and neuronal dynamics in the human posterior parietal cortex during evidence accumulation. Neuroimage 107, 219-228. doi: 10.1016/j.neuroimage.2014.12.015
  23. Fogassi, L., Ferrari, P. F., Gesierich, B., Rozzi, S., Chersi, F., and Rizzolotti, G. (2005). Parietal lobe: from action organization to intention understanding. Science 308, 662-667. doi: 10.1126/science.1106138
  24. Franklin, D. W., and Wolpert, D. M. (2011). Computational mechanisms of sensorimotor control. Neuron 72, 425-442. doi: 10.1016/j.neuron.2011.10.006
  25. Friston, K. (2008). Hierarchical models in the brain. PLoS Comput. Biol. 4, e1000211. doi: 10.1371/journal.pcbi.1000211
  26. Friston, K. (2010). The free-energy principle: a unified brain theory? Nat. Rev. Neurosci. 11, 127-138. doi: 10.1038/nrn2787
  27. Friston, K. (2011). What is optimal about motor control? Neuron 72, 488-498. doi: 10.1016/j.neuron.2011.10.018
  28. Friston, K. (2012). The history of the future of the Bayesian brain. Neuroimage 62, 1230-1233. doi: 10.1016/j.neuroimage.2011.10.004
  29. Friston, K., and Kiebel, S. (2009). Predictive coding under the free-energy principle. Philos. Trans. R. Soc. B Biol. Sci. 364, 1211-1221. doi: 10.1098/rstb.20
  30. Friston, K., Mattout, J., Trujillo-Barreto, N., Ashburner, J., and Penny, W. (2007). Variational free energy and the Laplace approximation. Neuroimage 34, 220-234. doi: 10.1016/j.neuroimage.2006.08.035
  31. Friston, K. J. (2002). Functional integration and inference in the brain. Progr. Neurobiol. 68, 113-143. doi: 10.1016/S0301-0082(02)00076-X Friston, K. J. (2005). A theory of cortical responses. Philos. Trans. R. Soc. Lond B Biol. Sci. 360, 815-836. doi: 10.1098/rstb.2005.1622
  32. Friston, K. J., Daunizeau, J., and Kiebel, S. J. (2009). Reinforcement learning or active inference? PLoS ONE 4, e6421. doi: 10.1371/journal.pone.
  33. Friston, K. J., Daunizeau, J., Kilner, J., and Kiebel, S. J. (2010). Action and behavior: a free-energy formulation. Biol. Cybern. 102, 227-260. doi: 10.1007/s00422-010-0364-z Friston, K. J., Mattout, J., and Kilner, J. (2011). Action understanding and active inference. Biol. Cybern. 104, 137-160. doi: 10.1007/s00422-011-0424-z Friston, K. J., Parr, T., and de Vries, B. (2017a). The graphical brain: belief propagation and active inference. Netw. Neurosci. 1, 381-414. doi: 10.1162/NETN_a_00018
  34. Friston, K. J., Rosch, R., Parr, T., Price, C., and Bowman, H. (2017b). Deep temporal models and active inference. Neurosci. Biobehav. Rev. 77, 388-402. doi: 10.1016/j.neubiorev.2017.04.009
  35. Friston, K. J., Samothrakis, S., and Montague, R. (2012). Active inference and agency: optimal control without cost functions. Biol. Cybern. 106, 523-541. doi: 10.1007/s00422-012-0512-8
  36. Friston, K. J., Trujillo-Barreto, N., and Daunizeau, J. (2008). DEM: A variational treatment of dynamic systems. Neuroimage 41, 849-885. doi: 10.1016/j.neuroimage.2008.02.054
  37. Gallego, J. A., Makin, T. R., and McDougle, S. D. (2022). Going beyond primary motor cortex to improve brain-computer interfaces. Trends Neurosci. 45, 176-183. doi: 10.1016/j.tins.2021.12.006
  38. Galletti, C., and Fattori, P. (2018). The dorsal visual stream revisited: Stable circuits or dynamic pathways? Cortex 98, 203-217. doi: 10.1016/j.cortex.2017.01.009
  39. Galletti, C., Gamberini, M., and Fattori, P. (2022). The posterior parietal area V6A: an attentionally-modulated visuomotor region involved in the control of reach-to-grasp action. Neurosci. Biobehav. Rev. 141, 104823. doi: 10.1016/j.neubiorev.2022.104823
  40. Gamberini, M., Passarelli, L., Filippini, M., Fattori, P., and Galletti, C. (2021). Vision for action: thalamic and cortical inputs to the macaque superior parietal lobule. Brain Struct. Funct. 226, 2951-2966. doi: 10.1007/s00429-021-02377-7
  41. Genovesio, A., Tsujimoto, S., and Wise, S. P. (2012). Encoding goals but not abstract magnitude in the primate prefrontal cortex. Neuron 74, 656-662. doi: 10.1016/j.neuron.2012.02.023
  42. Goodfellow, I. J., Bengio, Y., and Courville, A. (2016). Deep Learning. Cambridge, MA: MIT Press.
  43. Haar, S., and Donchin, O. (2020). A revised computational neuroanatomy for motor control. J. Cogn. Neurosci. 32, 1823-1836. doi: 10.1162/jocn_a_01602
  44. Hohwy, J. (2013). The Predictive Mind. Oxford: Oxford University Press UK. doi: 10.1093/acprof:oso/9780199682737.001.0001
  45. Kaplan, R., and Friston, K. J. (2018). Planning and navigation as active inference. Biol. Cybern. 112, 323-343. doi: 10.1007/s00422-018-0753-2
  46. Keele, S. W., and Posner, M. I. (1968). Processing of visual feedback in rapid movements. J. Exp. Psychol. 77, 155-158. doi: 10.1037/h0025754
  47. Kikuchi, Y., and Hamada, Y. (2009). Geometric characters of the radius and tibia in Macaca mulatta and Macaca fascicularis. Primates 50, 169-183. doi: 10.1007/s10329-008-0120-3
  48. Kingma, D. P., and Welling, M. (2014). "Auto-encoding variational bayes, " in 2nd International Conference on Learning Representations, ICLR 2014-Conference Track Proceedings (Banff), 1-14. doi: 10.48550/arXiv.1312.6114
  49. Kornblum, S., Hasbroucq, T., and Osman, A. (1990). Dimensional overlap: cognitive basis for stimulus-response compatibility-a model and taxonomy. Psychol. Rev. 97, 253-270. doi: 10.1037/0033-295X.97.2.253
  50. Lanillos, P., and Cheng, G. (2018). "Adaptive robot body learning and estimation through predictive coding, " in IEEE International Conference on Intelligent Robots and Systems (Madrid: IEEE), 4083-4090.
  51. Lanillos, P., Pages, J., and Cheng, G. (2020). "Robot self/other distinction: active inference meets neural networks learning in a mirror, " in ECAI 2020 (Santiago de Compostela). doi: 10.48550/arXiv.2004.05473
  52. Lau, H. C., Rogers, R. D., Haggard, P., and Passingham, R. E. (2004). Attention to Intention. Sicence 303, 1208-1210. doi: 10.1126/science.1090973
  53. Levine, S. (2018). Reinforcement learning and control as probabilistic inference: tutorial and review. ArXiv [Preprint]. doi: 10.48550/arXiv.1805.00909
  54. Limanowski, J., and Friston, K. (2020). Active inference under visuo- proprioceptive conflict: simulation and empirical results. Sci. Rep. 10, 1-14. doi: 10.1038/s41598-020-61097-w Ma, W. J., Beck, J. M., Latham, P. E., and Pouget, A. (2006). Bayesian inference with probabilistic population codes. Nat. Neurosci. 9, 1432-1438. doi: 10.1038/nn1790
  55. Medendorp, W. P., and Heed, T. (2019). State estimation in posterior parietal cortex: distinct poles of environmental and bodily states. Progr. Neurobiol. 183, 101691. doi: 10.1016/j.pneurobio.2019.101691
  56. Millidge, B., Tschantz, A., Seth, A. K., and Buckley, C. L. (2020). On the relationship between active inference and control as inference. Commun. Comput. Inf. Sci. 1326, 3-11. doi: 10.1007/978-3-030-64919-7_1
  57. Oliver, G., Lanillos, P., and Cheng, G. (2019). Active inference body perception and action for humanoid robots. ArXiv [Preprint]. doi: 10.48550/arXiv.1906.03022
  58. Parr, T., and Friston, K. J. (2018). The anatomy of inference: Generative models and brain structure. Front. Comput. Neurosci. 12, 90. doi: 10.3389/fncom.2018.00090
  59. Parr, T., Pezzulo, G., and Friston, K. J. (2022). Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. Cambridge, MA: The MIT Press. doi: 10.7551/mitpress/12441.001.0001
  60. Parr, T., Rikhye, R. V., Halassa, M. M., and Friston, K. J. (2020). Prefrontal computation as active inference. Cereb. Cortex 30, 682-695. doi: 10.1093/cercor/bhz118
  61. Pezzulo, G., and Cisek, P. (2016). Navigating the affordance landscape: feedback control as a process model of behavior and cognition. Trends Cogn. Sci. 20, 414-424. doi: 10.1016/j.tics.2016.03.013
  62. Pezzulo, G., Donnarumma, F., Dindo, H., D'Ausilio, A., Konvalinka, I., and Castelfranchi, C. (2019). The body talks: sensorimotor communication and its brain and kinematic signatures. Phys. Life Rev. 28, 1-21. doi: 10.1016/j.plrev.2018.06.014
  63. Pezzulo, G., Donnarumma, F., Iodice, P., Maisto, D., and Stoianov, I. (2017).
  64. Model-based approaches to active perception and control. Entropy 19, 266. doi: 10.3390/e19060266
  65. Pezzulo, G., Rigoli, F., and Friston, K. J. (2018). Hierarchical active inference: a theory of motivated control. Trends Cogn. Sci. 22, 294-306. doi: 10.1016/j.tics.2018.01.009
  66. Pio-Lopez, L., Nizard, A., Friston, K., and Pezzulo, G. (2016). Active inference and robot control: a case study. J. R. Soc. Interface 13, 122. doi: 10.1098/rsif.2016.0616
  67. Rao, R. P., and Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat. Neurosci. 2, 79-87. doi: 10.1038/4580
  68. Rood, T., van Gerven, M., and Lanillos, P. (2020). "A deep active inference model of the rubber-hand illusion, " in Active Inference. IWAI 2020. Communications in Computer and Information Science, Vol. 1326, eds T. Verbelen, P. Lanillos, C. L. Buckley and C. De Boom (Cham: Springer).
  69. Sajid, N., Ball, P. J., Parr, T., and Friston, K. J. (2021). Active inference: demystified and compared. Neural Comput. 33, 674-712. doi: 10.1162/neco_a_01357
  70. Sancaktar, C., van Gerven, M. A. J., and Lanillos, P. (2020). "End-to-end pixel- based deep active inference for body perception and action, " in 2020 Joint IEEE 10th International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob) (Valparaiso: IEEE), 1-8.
  71. Saunders, J. A., and Knill, D. C. (2003). Humans use continuous visual feedback from the hand to control fast reaching movements. Exp. Brain Res. 152, 341-352. doi: 10.1007/s00221-003-1525-2
  72. Shadmehr, R., and Krakauer, J. W. (2008). A computational neuroanatomy for motor control. Exp. Brain Res. 185, 359-381. doi: 10.1007/s00221-008-1280-5
  73. Shenoy, K. V., Sahani, M., and Churchland, M. M. (2013). Cortical control of arm movements: a dynamical systems perspective. Annu. Rev. Neurosci. 36, 337-359. doi: 10.1146/annurev-neuro-062111-150509
  74. Snyder, L. H., Batista, A. P., and Andersen, R. A. (1997). Coding of intention in the posterior parietal cortex. Nature 386, 167-170. doi: 10.1038/386167a0
  75. Snyder, L. H., Batista, A. P., and Andersen, R. A. (2000). Intention-related activity in the posterior parietal cortex: a review. Vision Res. 40, 1433-1441. doi: 10.1016/S0042-6989(00)00052-3
  76. Srinivasan, S. S., Gutierrez-Arango, S., Teng, A. C. E., Israel, E., Song, H., Bailey, Z. K., et al. (2021). Neural interfacing architecture enables enhanced motor control and residual limb functionality postamputation. Proc. Natl. Acad. Sci. U.S.A. 118, e2019555118. doi: 10.1073/pnas.2019555118
  77. Stoianov, I., Genovesio, A., and Pezzulo, G. (2016). Prefrontal goal codes emerge as latent states in probabilistic value learning. J. Cogn. Neurosci. 28, 140-157. doi: 10.1162/jocn_a_00886
  78. Stoianov, I., Kramer, P., Umiltà, C., and Zorzi, M. (2008). Visuospatial priming of the mental number line. Cognition. 106, 770-779. doi: 10.1016/j.cognition.2007.04.013
  79. Stoianov, I., Maisto, D., and Pezzulo, G. (2022). The hippocampal formation as a hierarchical generative model supporting generative replay and continual learning. Progr. Neurobiol. 217, 1-20. doi: 10.1016/j.pneurobio.2022. 102329
  80. Stoianov, I., Pennartz, C., Lansink, C., and Pezzulo, G. (2018). Model- based spatial navigation in the hippocampus-ventral striatum circuit: a computational analysis. PLoS Comput. Biol. 14, 1-28. doi: 10.1371/journal.pcbi. 1006316
  81. Stoianov, I., and Zorzi, M. (2012). Emergence of a 'visual number sense' in hierarchical generative models. Nat. Neurosci. 15, 194-196. doi: 10.1038/nn.2996
  82. Todorov, E. (2004). Optimality principles in sensorimotor control. Nat. Neurosci. 7, 907-915. doi: 10.1038/nn1309
  83. Todorov, E., and Jordan, M. I. (2002). Optimal feedback control as a theory of motor coordination. Nat. Neurosci. 5, 1226-1235. doi: 10.1038/nn963
  84. Toussaint, M., and Storkey, A. (2006). Probabilistic inference for solving discrete and continuous state Markov Decision Processes. ACM Int. Conf. Proceed. Ser. 148, 945-952. doi: 10.1145/1143844.1143963
  85. Tuthill, J. C., and Azim, E. (2018). Proprioception. Curr. Biol. 28, R194-R203. doi: 10.1016/j.cub.2018.01.064
  86. Velliste, M., Perel, S., Spalding, M. C., Whitford, A. S., and Schwartz, A. B. (2008). Cortical control of a prosthetic arm for self-feeding. Nature 453, 1098-1101. doi: 10.1038/nature06996
  87. Versteeg, C., Rosenow, J. M., Bensmaia, S. J., and Miller, L. E. (2021). Encoding of limb state by single neurons in the cuneate nucleus of awake monkeys. J. Neurophysiol. 126, 693-706. doi: 10.1152/jn.00568.2020
  88. Wolpert, D. M., and Flanagan, J. R. (2016). Computations underlying sensorimotor learning. Curr. Opin. Neurobiol. 37, 7-11. doi: 10.1016/j.conb.2015.12.003

Related papers

Decoding Action Intentions in Parietofrontal Circuits
Marco Davare

Journal of Neuroscience, 2011

View PDFchevron_right
A Neural Active Inference Model of Perceptual-Motor Learning
Gabriel Diaz

arXiv (Cornell University), 2022

The active inference framework (AIF) is a promising new computational framework grounded in contemporary neuroscience that can produce human-like behavior through reward-based learning. In this study, we test the ability for the AIF to capture the role of anticipation in the visual guidance of action in humans through the systematic investigation of a visual-motor task that has been well-explored -that of intercepting a target moving over a ground plane. Previous research demonstrated that humans performing this task resorted to anticipatory changes in speed intended to compensate for semi-predictable changes in target speed later in the approach. To capture this behavior, our proposed "neural" AIF agent uses artificial neural networks to select actions on the basis of a very short term prediction of the information about the task environment that these actions would reveal along with a long-term estimate of the resulting cumulative expected free energy. Systematic variation revealed that anticipatory behavior emerged only when required by limitations on the agent's movement capabilities, and only when the agent was able to estimate accumulated free energy over sufficiently long durations into the future. In addition, we present a novel formulation of the prior function that maps a multi-dimensional world-state to a uni-dimensional distribution of free-energy. Together, these results demonstrate the use of AIF as a plausible model of anticipatory visually guided behavior in humans.

View PDFchevron_right
Motor intention in the posterior parietal cortex
Jugoslava Acimovic

2008

The complexity of processes occurring in the brain is an intriguing issue not just for scientists and medical doctors, but the humanity in general. The cortex ability to perceive and analyze an enormous amount of information in an instance of time, the parallelism and computational efficiency are among the questions that attract attention. Even a simple, everyday gesture, for example, reaching for a cup of coffee, evokes a flow of signals in the brain. It goes from the primary visual region, that locates the cup on the table, to the primary motor region that sends the precise coordinates to the hand, and the instruction what to do next. The sequence of signal transmission and transformation continues through several regions, sensory, associative and motor ones. In this study, we will focus on the posterior parietal cortex, the region involved in the transformation of visual inputs into the preliminary motor plans. Les résultats obtenus démontrent que l'intention motrice peut être extraite des signaux neuronaux avec une précision d'environ 70% pour un groupe de cellules sélectionnées aléatoirement. Pour les meilleurs groupes de cellules, la précision atteint 82%. L'identification d'intentions motrices est particulièrement précise pendant les phases de préparation et la réalisation de mouvements oculaires. Cela est cohérent avec la fonction connue de la région 7a, où la majorité des cellules répondent aux mouvements oculaires. La précision est déterminée par la taille de la population considérée : la précision augmente avec une plus grande population. Néanmoins cette conclusion n'est valable qu'en général : ajouter une ou plusieurs cellules sélectionnées aléatoirement ne change pas nécessairement le résultat, car ces cellules peuvent ne pas apporter d'informations supplémentaires. L'influence de v chaque cellule présente dans un groupe est testée dans ce contexte. Les résultats obtenus montrent que le codage de l'intention motrice dans le cortex pariétal est redondant. Plusieurs cellules portent la même information et certaines d'entre elles peuvent être retirée du groupe sans que la précision de l'algorithme change. Néanmoins toutes les retirer dégrade le résultat. Finalement, l'influence de la taille de la fenêtre utilisée pour le calcul des impulsions est étudiée. En général, la précision augmente lorsque l'on utilise des fenêtres plus larges, résultat cohérent avec la littérature. Finalement, l'analyse des données expérimentales est complétée par une étude d'un modèle approprié. Modéliser des processus biologiques, afin de trouver des fonctions additionnelles et tester des paramètres non-accessibles à travers les données, est une approche largement acceptée. Néanmoins, le développement d'un modèle, suffisamment simple pour être implémenté sur les machines actuelles, mais suffisamment complexe pour capturer les processus d'intérêts, n'est pas évident. Notre motivation pour valider cette approche a été de tester plusieurs paramètres qui s'imposent comme importants dans les étapes d'analyse de données. Dû aux limitations des expériences, ainsi qu'à la nature du problème, les tests sur le modèle, même approximatif, étaient la seule tactique possible. L'influence de la taille de la population et le taille de la fenêtre a été caractérisée dans cette étude. Cela démontre que la précision de l'algorithme est fonction du nombre de cellules. Mots-clés: apprentissage automatique, intention motrice, codage de population, coordination main-oeil, cortex pariétal postérieur, fréquence d'impulsions, interfaces cerveau-machine, machines à état liquide, modèles bio-inspirés, paramètres moteurs, région 7a, réseaux récurrents, signal neurologique, train d'impulsions. vi I would like to express gratitude to my thesis adviser, Prof. Martin Hasler, for his support and patience during the four years that I spent in this laboratory. Many things tat I learn about science and life will follow me for all my life. Also, I would like to thank to my collaborators on this project, Prof. Alexandra Battaglia-Mayer and Prof. Roberto Caminiti from the University of Rome, whose experimental work on the parietal cortex functions served as the inspiration for this whole project. I am very grateful for their numerous explanations of the experiments, data, and in general, neurophysiologic aspects of this project, and also for their hospitality during my visits to Rome. Mr. Alessandro Palladini, from the University of Bologna, had an important contribution to development of the artificial model presented here. His interest in this project, and his positive and dynamic approach to the work, in general, made this part of the project possible. I had a great pleasure to spend some time outside of EPFL, on the internships in Belgium and Finland, during the second and the fourth year of my doctoral studies. These travels gave me the opportunity to look at my work from the new perspectives, but also, to learn a lot about other countries and cultures. I would like to thank to Prof. Johan Suykens from KU Leuven, and Dr. Marja-Leena Linne and Prof. Olli Yli-Harja from Tampere University of Technology, for their hospitality and help. Finally, I would like to thank to the members of my thesis committee, Prof. Jean-Yves Le Boudec, the president of the jury, and Dr. Marja-Leena Linne, Prof. Giorgio Innocenti and Prof. Wulfram Gerstner, the committee members. Thank you for reading and correcting my thesis, and for the very interesting discussion of the results during my defense. I really appreciated the atmosphere in the laboratory and the communication among lab members. It was great to work in a friendly and supportive environment, which I will always remember. Particularly, I would like to thank to Kumiko, who shared the office with me during all these years, and whose patience and calmness helped me to deal better with all the challenges of doctoral studies. I would like to thank to my parents and my brother for their love and care, and their unconditional support and understanding in many difficult moments of my live. I know how difficult for you was all this time that we spent far from each others. I was very fortunate to meet many wonderful people since coming to Lausanne, and make some friendships for life. If every new friend has made me a little bit better person, then, I have already gained a lot from coming to Lausanne. Many thanks to Zorana, my best friend and roommate during the last three years. And Stane, I actually do not know how to thank you. I do not think it is possible to express it in words, how thankful I am for having you in my life, how important you are to me. I would not manage to finish this without you. vii Contents Acknowledgments vii 1 Introduction 2 Neurophysiological background and motivation: The functional role of the posterior parietal cortex.

View PDFchevron_right
How the brain understands intention: Different neural circuits identify the componential features of motor and prior intentions
Cristina Becchio

Consciousness and Cognition, 2006

In this paper we present theoretical and experimental evidence for a set of mechanisms by which intention is understood. We propose that three basic aspects are involved in the understanding of intention. The first aspect to consider is intention recognition, i.e., the process by which we recognize other peopleÕs intentions, distinguishing among different types. The second aspect concerns the attribution of intention to its author: the existence of shared neural representations provides a parsimonious explanation of how we recognize other peopleÕs intentions (what they are doing), but in and of itself, is not sufficient to determine who the agent is. Once the intention has been recognized and attributed to an agent, the reasons for, and the aim of, the intention are to be considered. Hence, the third aspect concerns the aim-intention motivating the execution of a certain action. We discuss the neural basis of these three theoretical aspects suggesting a conceptual synthesis.

View PDFchevron_right
A Model of Intention Understanding Based on Learned Chains of Motor Acts in the Parietal Lobe
Albert Mukovskiy

2006

Understanding the actions of others is a fundamental cognitive skill for all individuals living in social groups. However, the underling neural mechanisms remain mostly unclear and thus difficult to implement on artificial systems. Recent neurophysiological experiments have revealed the existence in monkey and human brain of a particular type of neurons called mirror neurons that are active both when the individual executes an action and observes the same action done by another individual. Moreover it has been shown that neurons in the monkey's inferior parietal lobule, while coding a specific motor act, show markedly different activation patterns depending upon the final goal of the action sequence in which the act is embedded. In this paper, we present a dynamic network model which represents in its layers the functionality of neurons in different interconnected brain areas known to be involved in action observation/execution tasks. The model aims at substantiating the idea that action understanding is a continuous process which combines sensory evidence, prior task knowledge and a goal-directed matching of action observation and action execution. We also propose and test a biologically plausible learning scheme which allows establishing during practice a goal-directed organization of the distributed network.

View PDFchevron_right

Related topics

  • Computer Science
    • 580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved