Computational psychiatry as a bridge from neuroscience to clinical applications

Nature Neuroscience, Год журнала: 2016, Номер 19(3), С. 404 - 413

Опубликована: Фев. 23, 2016

Язык: Английский

---

How Do Expectations Shape Perception?

Trends in Cognitive Sciences, Год журнала: 2018, Номер 22(9), С. 764 - 779

Опубликована: Июнь 29, 2018

Язык: Английский

---

The Predictive Coding Account of Psychosis

Biological Psychiatry, Год журнала: 2018, Номер 84(9), С. 634 - 643

Опубликована: Май 25, 2018

Fueled by developments in computational neuroscience, there has been increasing interest the underlying neurocomputational mechanisms of psychosis. One successful approach involves predictive coding and Bayesian inference. Here, inferences regarding current state world are made combining prior beliefs with incoming sensory signals. Mismatches between signals constitute prediction errors that drive new learning. Psychosis suggested to result from a decreased precision encoding relative data, thereby garnering maladaptive inferences. we review evidence for aberrant discuss challenges this canonical account For example, hallucinations delusions may relate distinct alterations coding, despite their common co-occurrence. More broadly, some studies implicate weakened psychosis, others find stronger priors. These might be answered more nuanced view coding. Different priors specified different modalities integration, deficits each modality need not uniform. Furthermore, hierarchical organization critical. Altered processes at lower levels hierarchy linearly related higher (and vice versa). Finally, theories do highlight active inference—the process through which effects our actions on sensations anticipated minimized. It is possible conflicting findings reconciled considering these complexities, portending framework psychosis equipped deal its many manifestations.

Язык: Английский

---

Active inference and learning

Neuroscience & Biobehavioral Reviews, Год журнала: 2016, Номер 68, С. 862 - 879

Опубликована: Июль 1, 2016

This paper offers an active inference account of choice behaviour and learning. It focuses on the distinction between goal-directed habitual how they contextualise each other. We show that habits emerge naturally (and autodidactically) from sequential policy optimisation when agents are equipped with state-action policies. In inference, has explorative (epistemic) exploitative (pragmatic) aspects sensitive to ambiguity risk respectively, where epistemic (ambiguity-resolving) enables pragmatic (reward-seeking) subsequent emergence habits. Although policies usually associated model-based model-free schemes, we find more important is belief-free belief-based schemes. The underlying (variational) belief updating provides a comprehensive (if metaphorical) process theory for several phenomena, including transfer dopamine responses, reversal learning, habit formation devaluation. Finally, reduces classical (Bellman) scheme, in absence ambiguity.

Язык: Английский

---

Dynamic Network Communication as a Unifying Neural Basis for Cognition, Development, Aging, and Disease

Biological Psychiatry, Год журнала: 2015, Номер 77(12), С. 1089 - 1097

Опубликована: Май 2, 2015

Perception, cognition, and social interaction depend upon coordinated neural activity. This coordination operates within noisy, overlapping, distributed networks operating at multiple timescales. These are built a structural scaffolding with intrinsic neuroplasticity that changes development, aging, disease, personal experience. In this article, we begin from the perspective successful interregional communication relies transient synchronization between distinct low-frequency (<80 Hz) oscillations, allowing for brief windows of via phase-coordinated local neuronal spiking. From this, construct theoretical framework dynamic network communication, arguing these reflect balance oscillatory coupling population spiking activity two levels interact. We theorize when is too strong, spike timing becomes synchronous; weak, disorganized. Each results in specific disruptions to communication. alterations dynamics may underlie cognitive associated healthy development addition neurological psychiatric disorders. A number disorders—including Parkinson's autism, depression, schizophrenia, anxiety—are abnormalities Although experience differ biological gray or white matter, neurotransmission, gene expression, our suggests any resultant behavioral normal disordered states their treatment product how physical processes affect

Язык: Английский

---

A computational psychiatry approach identifies how alpha-2A noradrenergic agonist Guanfacine affects feature-based reinforcement learning in the macaque

Scientific Reports, Год журнала: 2017, Номер 7(1)

Опубликована: Янв. 16, 2017

Abstract Noradrenaline is believed to support cognitive flexibility through the alpha 2A noradrenergic receptor (a2A-NAR) acting in prefrontal cortex. Enhanced has been inferred from improved working memory with a2A-NA agonist Guanfacine. But it unclear whether Guanfacine improves specific attention and learning mechanisms beyond memory, drug effects can be formalized computationally allow single subject predictions. We tested confirmed these suggestions a case study healthy nonhuman primate performing feature-based reversal task evaluating performance using Bayesian Reinforcement models. In an initial dose-testing phase we found dose that increased accuracy, decreased distractibility learning. second experimental only examined faster single-subject computational modeling. Parameter estimation suggested not accounted for by varying reinforcement mechanism, but changing set of parameter values higher rates stronger suppression non-chosen over chosen feature information. These findings provide important starting point developing models discern synaptic functions within context neuropsychiatry framework.

Язык: Английский

---

Mood as Representation of Momentum

Trends in Cognitive Sciences, Год журнала: 2015, Номер 20(1), С. 15 - 24

Опубликована: Ноя. 4, 2015

TrendsWith increasing use of computational models to understand human behavior, scientists have begun model the dynamics subjective states such as mood.Recent data suggest that mood reflects cumulative impact differences between reward outcomes and expectations.Behavioral neural findings biases perception are perceived better when one is in a good relative bad mood.These two lines research establish bidirectional interaction reinforcement learning, which may play an important adaptive role healthy whose dysfunction might contribute psychiatric disorders.AbstractExperiences affect mood, turn affects subsequent experiences. Recent studies specific principles. First, depends on how recent differ from expectations. Second, way we perceive (e.g., rewards), this bias learning about those outcomes. We propose two-way serves mitigate inefficiencies application real-world problems. Specifically, represents overall momentum outcomes, its biasing influence 'corrects' account for environmental dependencies. describe potential dysfunctions mechanism symptoms disorders.

Язык: Английский

---

Revealing Neurocomputational Mechanisms of Reinforcement Learning and Decision-Making With the hBayesDM Package

Computational Psychiatry, Год журнала: 2017, Номер 1(0), С. 24 - 24

Опубликована: Авг. 25, 2017

Reinforcement learning and decision-making (RLDM) provide a quantitative framework computational theories with which we can disentangle psychiatric conditions into the basic dimensions of neurocognitive functioning. RLDM offer novel approach to assessing potentially diagnosing patients, there is growing enthusiasm for both psychiatry among clinical researchers. Such also insights brain substrates particular processes, as exemplified by model-based analysis data from functional magnetic resonance imaging (fMRI) or electroencephalography (EEG). However, researchers often find too technical have difficulty adopting it their research. Thus, critical need remains develop user-friendly tool wide dissemination methods. We introduce an R package called hBayesDM (hierarchical Bayesian modeling underline;">Decision-Making tasks), offers array tasks social exchange games. The state-of-the-art hierarchical Bayesian modeling, in individual group parameters (i.e., posterior distributions) are estimated simultaneously mutually constraining fashion. At same time, extremely user-friendly: users perform output visualization, model comparisons, each single line coding. Users extract trial-by-trial latent variables (e.g., prediction errors) required fMRI/EEG. With package, anticipate that anyone minimal knowledge programming take advantage cutting-edge computational-modeling approaches investigate underlying processes interactions between multiple goal-directed, habitual, Pavlovian) systems. In this way, expect will contribute advanced enable range easily research within different populations.

Язык: Английский

---

Computations of uncertainty mediate acute stress responses in humans

Nature Communications, Год журнала: 2016, Номер 7(1)

Опубликована: Март 29, 2016

The effects of stress are frequently studied, yet its proximal causes remain unclear. Here we demonstrate that subjective estimates uncertainty predict the dynamics and physiological responses. Subjects learned a probabilistic mapping between visual stimuli electric shocks. Salivary cortisol confirmed our stressor elicited changes in endocrine activity. Using hierarchical Bayesian learning model, quantified relationship different forms task acute Subjective stress, pupil diameter skin conductance all tracked evolution irreducible uncertainty. We observed coupling emotional somatic state, with tuning to tightly correlated. Furthermore, predicted individual performance, consistent an adaptive role for under uncertain threat. Our finding responses tuned environmental provides new insight into their generation likely function.

Язык: Английский

---

Computational Psychiatry: towards a mathematically informed understanding of mental illness

Journal of Neurology Neurosurgery & Psychiatry, Год журнала: 2015, Номер unknown, С. jnnp - 310737

Опубликована: Июль 8, 2015

Computational Psychiatry aims to describe the relationship between brain9s neurobiology, its environment and mental symptoms in computational terms. In so doing, it may improve psychiatric classification diagnosis treatment of illness. It can unite many levels description a mechanistic rigorous fashion, while avoiding biological reductionism artificial categorisation. We how models cognition infer current state weigh up future actions, these provide new perspectives on two example disorders, depression schizophrenia. Reinforcement learning describes brain choose value courses actions according their long-term value. Some depressive result from aberrant valuations, which could arise prior beliefs about loss agency (‘helplessness’), or an inability inhibit exploration aversive events. Predictive coding explains might perform Bayesian inference by combining sensory data with beliefs, each weighted certainty (or precision). Several cortical abnormalities schizophrenia reduce precision at higher inferential hierarchy, biasing towards away beliefs. discuss whether striatal hyperdopaminergia have adaptive function this context, also reinforcement incentive salience shed light disorder. Finally, we review some Psychiatry9s applications neurological such as Parkinson9s disease, pitfalls avoid when applying methods.

Язык: Английский

---