Deep temporal models and active inference

Friston, Karl J.; Rosch, Richard; Parr, Thomas; Price, Cathy J.; Bowman, Howard

doi:10.1016/j.neubiorev.2017.04.009

Cited by 191 publications

(160 citation statements)

References 83 publications

Supporting

Mentioning

156

Contrasting

Order By: Relevance

“…It involves an anticipation of mental agency and the capacity for global self-control (Wiese, 2019), the subjective but as yet non-egoic confidence that a full-blown first-person perspective involving executive control on the mental level plus an extended autobiographical self-model will very soon appear. Therefore, the phenomenal quality of subjective confidence not only relates to knowing • Spontaneity: Finally, there is a fifth global aspect characterizing the phenomenology corresponding to the functionally autonomous process of tonic alertness as bare, apparently self-generating and apparently ever-fresh wakeful awareness, and it is related to the absence of time-representationto the "temporal thinness" mentioned above (Friston et al, 2017). MPE has a specific quality of spontaneity, as it is ahistorical and contains not even an implicit representation of its own causal history.…”

Section: The Phenomenology Of Phasic Vs Tonic Alertnessmentioning

confidence: 99%

Minimal phenomenal experience

Metzinger¹

2020

PhiMiSci

109

150

View full text Add to dashboard Cite

This is the first in a series of instalments aiming at a minimal model explanation for conscious experience, taking the phenomenal character of “pure consciousness” or “pure awareness” in meditation as its entry point. It develops the concept of “minimal phenomenal experience” (MPE) as a candidate for the simplest form of consciousness, substantiating it by extracting six semantic constraints from the existing literature and using sixteen phenomenological case-studies to incrementally flesh out the new working concept. One empirical hypothesis is that the phenomenological prototype of “pure awareness”, to which all such reports refer, really is the content of a predictive model, namely, a Bayesian representation of tonic alertness. On a more abstract conceptual level, it can be described as a model of an unpartitioned epistemic space.

show abstract

Section: The Phenomenology Of Phasic Vs Tonic Alertnessmentioning

confidence: 99%

Minimal phenomenal experience

Metzinger¹

2020

PhiMiSci

109

150

View full text Add to dashboard Cite

show abstract

“…Such an epistemic drive has the effect of promoting actions that uncover information about hidden states via sampling informative observations. This intrinsic drive to uncover information, and its natural emergence via the minimization of expected free energy, is integral to accounts of exploratory behaviour, curiosity, salience, and related active-sensing phenomena under active inference (FitzGerald, Dolan, & Friston, 2015; Friston, Rosch, Parr, Price, & Bowman, 2017; Friston, Lin, et al, 2017; Parr & Friston, 2017b, 2018b; Mirza, Adams, Parr, & Friston, 2019). An alternative formulation of the expected free energy is given in the second line of Equation (10), where minimizing expected free energy promotes policies that reduce ‘ambiguity,’ defined as the expected uncertainty of observations, given the states expected under a policy.…”

Section: Free Energy Minimization and Active Inferencementioning

confidence: 99%

Deep Active Inference and Scene Construction

Heins

Mirza

Parr

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

14Adaptive agents must act in intrinsically uncertain environments with complex latent structure. Here, 15 we elaborate a model of visual foraging -in a hierarchical context -wherein agents infer a higher-order 16 visual pattern (a 'scene') by sequentially sampling ambiguous cues. Inspired by previous models of scene 17 construction -that cast perception and action as consequences of approximate Bayesian inference -we 18 use active inference to simulate decisions of agents categorizing a scene in a hierarchically-structured set-19 ting. Under active inference, agents develop probabilistic beliefs about their environment, while actively 20 sampling it to maximise the evidence for their internal generative model. This approximate evidence 21 maximization (i.e. self-evidencing) comprises drives to both maximise rewards and resolve uncertainty 22 about hidden states. This is realised via minimization of a free energy functional of posterior beliefs 23 about both the world as well as the actions used to sample or perturb it, corresponding to perception 24and action, respectively. We show that active inference, in the context of hierarchical scene construction, 25 gives rise to many empirical evidence accumulation phenomena, such as noise-sensitive reaction times 26 and epistemic saccades. We explain these behaviours in terms of the principled drives that constitute the 27 expected free energy, the key quantity for evaluating policies under active inference. In addition, we report 28 novel behaviours exhibited by these active inference agents that furnish new predictions for research on 29 evidence accumulation and perceptual decision-making. We discuss the implications of this hierarchical 30 active inference scheme for tasks that require planned sequences of information-gathering actions to infer 31 compositional latent structure (such as visual scene construction and sentence comprehension). Finally, 32 we propose experiments to contextualise active inference in relation to other formulations of evidence 33 accumulation (e.g. drift-diffusion models) in tasks that require planning in uncertain environments with 34 higher-order structure. 35

show abstract

“…Predictive mechanisms in neural network models are thought to involve an internal forward model (Wolpert & Miall, ) which has been often formalised in an information theoretical framework using Bayesian statistics as a generative model, generating (predicting) inputs from stored representations (Dayan & Abbott, ; Friston & Kiebel, ). Time is of special significance in the auditory sensory domain and one of the parameters that can influence predictive coding (Sedley et al ., ), although rarely examined in the context of predictive coding (but see Friston et al ., ).…”

Section: Introductionmentioning

confidence: 97%

Predictive cues for auditory stream formation in humans and monkeys

Aggelopoulos

Deike

Selezneva

et al. 2018

Eur J of Neuroscience

View full text Add to dashboard Cite

Auditory perception is improved when stimuli are predictable, and this effect is evident in a modulation of the activity of neurons in the auditory cortex as shown previously. Human listeners can better predict the presence of duration deviants embedded in stimulus streams with fixed interonset interval (isochrony) and repeated duration pattern (regularity), and neurons in the auditory cortex of macaque monkeys have stronger sustained responses in the 60-140 ms post-stimulus time window under these conditions. Subsequently, the question has arisen whether isochrony or regularity in the sensory input contributed to the enhancement of the neuronal and behavioural responses. Therefore, we varied the two factors isochrony and regularity independently and measured the ability of human subjects to detect deviants embedded in these sequences as well as measuring the responses of neurons the primary auditory cortex of macaque monkeys during presentations of the sequences. The performance of humans in detecting deviants was significantly increased by regularity. Isochrony enhanced detection only in the presence of the regularity cue. In monkeys, regularity increased the sustained component of neuronal tone responses in auditory cortex while isochrony had no consistent effect. Although both regularity and isochrony can be considered as parameters that would make a sequence of sounds more predictable, our results from the human and monkey experiments converge in that regularity has a greater influence on behavioural performance and neuronal responses.

show abstract

Deep temporal models and active inference

Cited by 191 publications

References 83 publications

Minimal phenomenal experience

Minimal phenomenal experience

Deep Active Inference and Scene Construction

Predictive cues for auditory stream formation in humans and monkeys

Contact Info

Product

Resources

About