This paper considers prediction and perceptual categorization as an inference problem that is solved by the brain. We assume that the brain models the world as a hierarchy or cascade of dynamical systems that encode causal structure in the sensorium. Perception is equated with the optimization or inversion of these internal models, to explain sensory data. Given a model of how sensory data are generated, we can invoke a generic approach to model inversion, based on a free energy bound on the model's evidence. The ensuing free-energy formulation furnishes equations that prescribe the process of recognition, i.e. the dynamics of neuronal activity that represent the causes of sensory input. Here, we focus on a very general model, whose hierarchical and dynamical structure enables simulated brains to recognize and predict trajectories or sequences of sensory states. We first review hierarchical dynamical models and their inversion. We then show that the brain has the necessary infrastructure to implement this inversion and illustrate this point using synthetic birds that can recognize and categorize birdsongs.Keywords: generative models; predictive coding; hierarchical; birdsong INTRODUCTIONThis paper reviews generic models of our sensorium and a Bayesian scheme for their inversion. We then show that the brain has the necessary anatomical and physiological equipment to invert these models, given sensory data. Critically, the scheme lends itself to a relatively simple neural network implementation that shares many features with real cortical hierarchies in the brain. The basic idea that the brain tries to infer the causes of sensations dates back to Helmholtz (e.g. Helmholtz 1860/1962Barlow 1961;Neisser 1967;Ballard et al. 1983;Mumford 1992;Kawato et al. 1993;Dayan et al. 1995;Rao & Ballard 1998), with a recent emphasis on hierarchical inference and empirical Bayes (Friston 2003(Friston , 2005Friston et al. 2006). Here, we generalize this idea to cover dynamics in the world and consider how neural networks could be configured to invert hierarchical dynamical models and deconvolve sensory causes from sensory input. This paper comprises four sections. In §1, we introduce hierarchical dynamical models and their inversion. These models cover most of the models encountered in the statistical literature. An important aspect of these models is their formulation in generalized coordinates of motion, which lends them a hierarchal form in both structure and dynamics. These hierarchies induce empirical priors that provide structural and dynamical constraints, which can be exploited during inversion. In §2, we show how inversion can be formulated as a simple gradient ascent using neuronal networks; in §3, we consider how evoked brain responses might be understood in terms of inference under hierarchical dynamical models of sensory input. 1

We have previously tried to explain perceptual inference and learning under a free-energy principle that pursues Helmholtz's agenda to understand the brain in terms of energy minimization. It is fairly easy to show that making inferences about the causes of sensory data can be cast as the minimization of a free-energy bound on the likelihood of sensory inputs, given an internal model of how they were caused. In this article, we consider what would happen if the data themselves were sampled to minimize this bound. It transpires that the ensuing active sampling or inference is mandated by ergodic arguments based on the very existence of adaptive agents. Furthermore, it accounts for many aspects of motor behavior; from retinal stabilization to goal-seeking. In particular, it suggests that motor control can be understood as fulfilling prior expectations about proprioceptiveThe free-energy principle is an attempt to explain the structure and function of the brain, starting from the fact that we exist: This fact places constraints on our interactions with the world, which have been studied for years in evolutionary biology and systems theory. However, recent advances in statistical physics and machine learning point to a simple scheme that enables biological systems to comply with these constraints. If one looks at the brain as implementing this scheme (minimizing a free-energy bound on disorder), then many aspects of its anatomy and physiology start to make sense. In this article, we show that free-energy can be reduced by selectively sampling sensory inputs. This leads to adaptive responses and provides a new view of how movement control might work in the brain. The main conclusion is that we only need to have expectations about the sensory consequences of moving in order to elicit movement. This means we that can replace the notion of desired movements with expected movements and understand action in terms of perceptual expectations.The Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, 12 Queen Square, London, WC1N 3BG, sensations. This formulation can explain why adaptive behavior emerges in biological agents and suggests a simple alternative to optimal control theory. We illustrate these points using simulations of oculomotor control and then apply to same principles to cued and goal-directed movements. In short, the free-energy formulation may provide an alternative perspective on the motor control that places it in an intimate relationship with perception.

In this paper, we suggest that cortical anatomy recapitulates the temporal hierarchy that is inherent in the dynamics of environmental states. Many aspects of brain function can be understood in terms of a hierarchy of temporal scales at which representations of the environment evolve. The lowest level of this hierarchy corresponds to fast fluctuations associated with sensory processing, whereas the highest levels encode slow contextual changes in the environment, under which faster representations unfold. First, we describe a mathematical model that exploits the temporal structure of fast sensory input to track the slower trajectories of their underlying causes. This model of sensory encoding or perceptual inference establishes a proof of concept that slowly changing neuronal states can encode the paths or trajectories of faster sensory states. We then review empirical evidence that suggests that a temporal hierarchy is recapitulated in the macroscopic organization of the cortex. This anatomic-temporal hierarchy provides a comprehensive framework for understanding cortical function: the specific time-scale that engages a cortical area can be inferred by its location along a rostro-caudal gradient, which reflects the anatomical distance from primary sensory areas. This is most evident in the prefrontal cortex, where complex functions can be explained as operations on representations of the environment that change slowly. The framework provides predictions about, and principled constraints on, cortical structure–function relationships, which can be tested by manipulating the time-scales of sensory input.

No abstract

Neuronally plausible, generative or forward models are essential for understanding how event-related fields (ERFs) and potentials (ERPs) are generated. In this paper, we present a new approach to modeling event-related responses measured with EEG or MEG. This approach uses a biologically informed model to make inferences about the underlying neuronal networks generating responses. The approach can be regarded as a neurobiologically constrained source reconstruction scheme, in which the parameters of the reconstruction have an explicit neuronal interpretation. Specifically, these parameters encode, among other things, the coupling among sources and how that coupling depends upon stimulus attributes or experimental context. The basic idea is to supplement conventional electromagnetic forward models, of how sources are expressed in measurement space, with a model of how source activity is generated by neuronal dynamics. A single inversion of this extended forward model enables inference about both the spatial deployment of sources and the underlying neuronal architecture generating them. Critically, this inference covers long-range connections among well-defined neuronal subpopulations. In a previous paper, we simulated ERPs using a hierarchical neural-mass model that embodied bottom-up, top-down and lateral connections among remote regions. In this paper, we describe a Bayesian procedure to estimate the parameters of this model using empirical data. We demonstrate this procedure by characterizing the role of changes in cortico-cortical coupling, in the genesis of ERPs. In the first experiment, ERPs recorded during the perception of faces and houses were modeled as distinct cortical sources in the ventral visual pathway. Category-selectivity, as indexed by the face-selective N170, could be explained by category-specific differences in forward connections from sensory to higher areas in the ventral stream. We were able to quantify and make inferences about these effects using conditional estimates of connectivity. This allowed us to identify where, in the processing stream, category-selectivity emerged. In the second experiment, we used an auditory oddball paradigm to show that the mismatch negativity can be explained by changes in connectivity. Specifically, using Bayesian model selection, we assessed changes in backward connections, above and beyond changes in forward connections. In accord with theoretical predictions, there was strong evidence for learning-related changes in both forward and backward coupling. These examples show that category- or context-specific coupling among cortical regions can be assessed explicitly, within a mechanistic, biologically motivated inference framework.

No abstract

Using dynamic causal modelling (DCM), we have presented provisional evidence to suggest: (i) the mismatch negativity (MMN) is generated by self-organised interactions within a hierarchy of cortical sources and (ii) the MMN rests on plastic change in both extrinsic (between-source) and intrinsic (within source) connections (Garrido et al., under review). In this work we re-visit these two key issues in the context of the roving paradigm. Critically, this paradigm allows us to discount any differential response to differences in the stimuli per se, because the standards and oddballs are physically identical. We were able to confirm both the hierarchical nature of the MMN generation and the conjoint role of changes in extrinsic and intrinsic connections. These findings are consistent with a predictive coding account of repetitionsuppression and the MMN, which gracefully accommodates two important mechanistic perspectives; the model adjustment hypothesis (Winkler et al., 1996;Näätänen and Winkler, 1999;Sussman and Winkler, 2001) and the adaptation hypothesis (May et al., 1999;and Jääskeläinen et al., 2004).

In Friston et al. ((2002)Neuroimage 16: 465-483) we introduced empirical Bayes as a potentially useful way to estimate and make inferences about effects in hierarchical models. In this paper we present a series of models that exemplify the diversity of problems that can be addressed within this framework. In hierarchical linear observation models, both classical and empirical Bayesian approaches can be framed in terms of covariance component estimation (e.g., variance partitioning). To illustrate the use of the expectationmaximization (EM) algorithm in covariance component estimation we focus first on two important problems in fMRI: nonsphericity induced by (i) serial or temporal correlations among errors and (ii) variance components caused by the hierarchical nature of multisubject studies. In hierarchical observation models, variance components at higher levels can be used as constraints on the parameter estimates of lower levels. This enables the use of parametric empirical Bayesian (PEB) estimators, as distinct from classical maximum likelihood (ML) estimates. We develop this distinction to address: (i) The difference between response estimates based on ML and the conditional means from a Bayesian approach and the implications for estimates of intersubject variability. (ii) The relationship between fixed-and random-effect analyses. (iii) The specificity and sensitivity of Bayesian inference and, finally, (iv) the relative importance of the number of scans and subjects. The forgoing is concerned with within-and between-subject variability in multisubject hierarchical fMRI studies. In the second half of this paper we turn to Bayesian inference at the first (within-voxel) level, using PET data to show how priors can be derived from the (between-voxel) distribution of activations over the brain. This application uses exactly the same ideas and formalism but, in this instance, the second level is provided by observations over voxels as opposed to subjects. The ensuing posterior probability maps (PPMs) have enhanced anatomical precision and greater face validity, in relation to underlying anatomy. Furthermore, in comparison to conventional SPMs they are not confounded by the multiple comparison problem that, in a classical context, dictates high thresholds and low sensitivity. We conclude with some general comments on Bayesian approaches to image analysis and on some unresolved issues. © 2002 Elsevier Science (USA)