1
|
Heins RC, Mirza MB, Parr T, Friston K, Kagan I, Pooresmaeili A. Deep Active Inference and Scene Construction. Front Artif Intell 2020; 3:509354. [PMID: 33733195 PMCID: PMC7861336 DOI: 10.3389/frai.2020.509354] [Citation(s) in RCA: 10] [Impact Index Per Article: 2.5] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/01/2019] [Accepted: 09/10/2020] [Indexed: 11/17/2022] Open
Abstract
Adaptive agents must act in intrinsically uncertain environments with complex latent structure. Here, we elaborate a model of visual foraging-in a hierarchical context-wherein agents infer a higher-order visual pattern (a "scene") by sequentially sampling ambiguous cues. Inspired by previous models of scene construction-that cast perception and action as consequences of approximate Bayesian inference-we use active inference to simulate decisions of agents categorizing a scene in a hierarchically-structured setting. Under active inference, agents develop probabilistic beliefs about their environment, while actively sampling it to maximize the evidence for their internal generative model. This approximate evidence maximization (i.e., self-evidencing) comprises drives to both maximize rewards and resolve uncertainty about hidden states. This is realized via minimization of a free energy functional of posterior beliefs about both the world as well as the actions used to sample or perturb it, corresponding to perception and action, respectively. We show that active inference, in the context of hierarchical scene construction, gives rise to many empirical evidence accumulation phenomena, such as noise-sensitive reaction times and epistemic saccades. We explain these behaviors in terms of the principled drives that constitute the expected free energy, the key quantity for evaluating policies under active inference. In addition, we report novel behaviors exhibited by these active inference agents that furnish new predictions for research on evidence accumulation and perceptual decision-making. We discuss the implications of this hierarchical active inference scheme for tasks that require planned sequences of information-gathering actions to infer compositional latent structure (such as visual scene construction and sentence comprehension). This work sets the stage for future experiments to investigate active inference in relation to other formulations of evidence accumulation (e.g., drift-diffusion models) in tasks that require planning in uncertain environments with higher-order structure.
Collapse
Affiliation(s)
- R. Conor Heins
- Department of Collective Behaviour, Max Planck Institute for Animal Behavior, Konstanz, Germany
- Perception and Cognition Group, European Neuroscience Institute, A Joint Initiative of the University Medical Centre Göttingen and the Max-Planck-Society, Göttingen, Germany
- Leibniz Science Campus “Primate Cognition”, Göttingen, Germany
| | - M. Berk Mirza
- Department of Neuroimaging, Institute of Psychiatry, Psychology and Neuroscience, King's College London, London, United Kingdom
- The National Institute for Health Research (NIHR) Maudsley Biomedical Research Centre (BRC) at South London and Maudsley National Health Service (NHS) Foundation Trust and The Institute of Psychiatry, Psychology and Neuroscience, King's College London, London, United Kingdom
- Wellcome Centre for Human Neuroimaging, University College London, London, United Kingdom
| | - Thomas Parr
- Wellcome Centre for Human Neuroimaging, University College London, London, United Kingdom
| | - Karl Friston
- Wellcome Centre for Human Neuroimaging, University College London, London, United Kingdom
| | - Igor Kagan
- Leibniz Science Campus “Primate Cognition”, Göttingen, Germany
- Decision and Awareness Group, Cognitive Neuroscience Laboratory, German Primate Centre (DPZ), Göttingen, Germany
| | - Arezoo Pooresmaeili
- Perception and Cognition Group, European Neuroscience Institute, A Joint Initiative of the University Medical Centre Göttingen and the Max-Planck-Society, Göttingen, Germany
- Leibniz Science Campus “Primate Cognition”, Göttingen, Germany
| |
Collapse
|
2
|
Juneau ZC, Stonemetz JM, Toma RF, Possidente DR, Heins RC, Vecsey CG. Optogenetic activation of short neuropeptide F (sNPF) neurons induces sleep in Drosophila melanogaster. Physiol Behav 2019; 206:143-156. [PMID: 30935941 PMCID: PMC6520144 DOI: 10.1016/j.physbeh.2019.03.027] [Citation(s) in RCA: 5] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/25/2018] [Revised: 01/18/2019] [Accepted: 03/28/2019] [Indexed: 01/31/2023]
Abstract
Sleep abnormalities have widespread and costly public health consequences, yet we have only a rudimentary understanding of the events occurring at the cellular level in the brain that regulate sleep. Several key signaling molecules that regulate sleep across taxa come from the family of neuropeptide transmitters. For example, in Drosophila melanogaster, the neuropeptide Y (NPY)-related transmitter short neuropeptide F (sNPF) appears to promote sleep. In this study, we utilized optogenetic activation of neuronal populations expressing sNPF to determine the causal effects of precisely timed activity in these cells on sleep behavior. Combining sNPF-GAL4 and UAS-Chrimson transgenes allowed us to activate sNPF neurons using red light. We found that activating sNPF neurons for as little as 3 s at a time of day when most flies were awake caused a rapid transition to sleep that persisted for another 2+ hours following the stimulation. Changing the timing of red light stimulation to times of day when flies were already asleep caused the control flies to wake up (due to the pulse of light), but the flies in which sNPF neurons were activated stayed asleep through the light pulse, and then showed further increases in sleep at later points when they would have normally been waking up. Video recording of individual fly responses to short-term (0.5-20 s) activation of sNPF neurons demonstrated a clear light duration-dependent decrease in movement during the subsequent 4-min period. These results provide supportive evidence that sNPF-producing neurons promote long-lasting increases in sleep, and show for the first time that even brief periods of activation of these neurons can cause changes in behavior that persist after cessation of activation. We have also presented evidence that sNPF neuron activation produces a homeostatic sleep drive that can be dissipated at times long after the neurons were stimulated. Future studies will determine the specific roles of sub-populations of sNPF-producing neurons, and will also assess how sNPF neurons act in concert with other neuronal circuits to control sleep.
Collapse
Affiliation(s)
- Zoe Claire Juneau
- Neuroscience Program, Skidmore College, 815 N. Broadway, Saratoga Springs, NY 12866, United States of America
| | - Jamie M Stonemetz
- Neuroscience Program, Skidmore College, 815 N. Broadway, Saratoga Springs, NY 12866, United States of America
| | - Ryan F Toma
- Neuroscience Program, Skidmore College, 815 N. Broadway, Saratoga Springs, NY 12866, United States of America
| | - Debra R Possidente
- Neuroscience Program, Skidmore College, 815 N. Broadway, Saratoga Springs, NY 12866, United States of America
| | - R Conor Heins
- Biology Department, Swarthmore College, 500 College Avenue, Swarthmore, PA 19081, United States of America
| | - Christopher G Vecsey
- Neuroscience Program, Skidmore College, 815 N. Broadway, Saratoga Springs, NY 12866, United States of America; Biology Department, Swarthmore College, 500 College Avenue, Swarthmore, PA 19081, United States of America.
| |
Collapse
|