Bayesian Computation through Cortical Latent Dynamics

• Published in: Neuron (Cambridge, Mass.) Vol. 103; no. 5; pp. 934 - 947.e5
• Main Authors: Sohn, Hansem, Narain, Devika, Meirhaeghe, Nicolas, Jazayeri, Mehrdad
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 04.09.2019; Elsevier Limited
• Subjects: Animals; Bayes Theorem; Bayesian analysis; Bayesian inference; Bayesian integration; Bias; Cerebral Cortex; Cognition; Cognition - physiology; Cortex (frontal); Experiments; frontal cortex; Frontal Lobe - physiology; Information processing; Macaca mulatta; Nerve Net - physiology; neural manifold; Neural networks; Neural Networks, Computer; neural trajectories; Neurons; Physiology; recurrent neural networks; Sensorimotor system; Somatosensory cortex; Statistical analysis; Time Perception; neural manifold; frontal cortex; neural trajectories; recurrent neural networks; Bayesian inference; Bayesian integration
• ISSN: 0896-6273; 1097-4199; 1097-4199
• DOI: 10.1016/j.neuron.2019.06.012

Statistical regularities in the environment create prior beliefs that we rely on to optimize our behavior when sensory information is uncertain. Bayesian theory formalizes how prior beliefs can be leveraged and has had a major impact on models of perception, sensorimotor function, and cognition. However, it is not known how recurrent interactions among neurons mediate Bayesian integration. By using a time-interval reproduction task in monkeys, we found that prior statistics warp neural representations in the frontal cortex, allowing the mapping of sensory inputs to motor outputs to incorporate prior statistics in accordance with Bayesian inference. Analysis of recurrent neural network models performing the task revealed that this warping was enabled by a low-dimensional curved manifold and allowed us to further probe the potential causal underpinnings of this computational strategy. These results uncover a simple and general principle whereby prior beliefs exert their influence on behavior by sculpting cortical latent dynamics. [Display omitted] •Monkeys estimate time by integrating sensory evidence with prior beliefs•Prior beliefs warp neural representations in the frontal cortex•Warped representations provide an optimal substrate for integrating beliefs•Recurrent neural network models validate the warping effect of prior beliefs Sohn et al. found that prior beliefs warp neural representations in the frontal cortex. This warping provides a substrate for the optimal integration of prior beliefs with sensory evidence during sensorimotor behavior.