Publications

Learning and Generalization under Ambiguity: An fMRI Study

Adaptive behavior often exploits generalizations from past experience by applying them judiciously in new situations. This requires a means of quantifying the relative importance of prior experience and current information, so they can be balanced optimally. In this study, we ask whether the brain generalizes in an optimal way. Specifically, we used Bayesian learning theory and fMRI to test whether neuronal responses reflect context-sensitive changes in ambiguity or uncertainty about experience-dependent beliefs. We found that the hippocampus expresses clear ambiguity-dependent responses that are associated with an augmented rate of learning. These findings suggest candidate neuronal systems that may be involved in aberrations of generalization, such as over-confidence.

Dynamic modeling of neuronal responses in fMRI using cubature Kalman filtering

Optimizing Data for Modeling Neuronal Responses

In this technical note, we address an unresolved challenge in neuroimaging statistics: how to determine which of several datasets is the best for inferring neuronal responses. Comparisons of this kind are important for experimenters when choosing an imaging protocol-and for developers of new acquisition methods. However, the hypothesis that one dataset is better than another cannot be tested using conventional statistics (based on likelihood ratios), as these require the data to be the same under each hypothesis. Here we present Bayesian data comparison (BDC), a principled framework for evaluating the quality of functional imaging data, in terms of the precision with which neuronal connectivity parameters can be estimated and competing models can be disambiguated. For each of several candidate datasets, neuronal responses are modeled using Bayesian (probabilistic) forward models, such as General Linear Models (GLMs) or Dynamic Casual Models (DCMs). Next, the parameters from subject-specific models are summarized at the group level using a Bayesian GLM. A series of measures, which we introduce here, are then used to evaluate each dataset in terms of the precision of (group-level) parameter estimates and the ability of the data to distinguish similar models. To exemplify the approach, we compared four datasets that were acquired in a study evaluating multiband fMRI acquisition schemes, and we used simulations to establish the face validity of the comparison measures. To enable people to reproduce these analyses using their own data and experimental paradigms, we provide general-purpose Matlab code via the SPM software.

Embodiment and Schizophrenia: A Review of Implications and Applications

In recent decades, embodiment has become an influential concept in psychology and cognitive neuroscience. Embodiment denotes the study of the reciprocal (causal) relationships between mind and body, with the mind not only affecting the body but also vice versa. Embodied cognition comes to the fore in sensorimotor coupling, predictive coding, and nonverbal behavior. Additionally, the embodiment of the mind constitutes the basis of social interaction and communication, as evident in research on nonverbal synchrony and mimicry. These theoretical and empirical developments portend a range of implications for schizophrenia research and treatment. Sensorimotor dysfunctions are closely associated with affective and psychotic psychopathology, leading to altered timing in the processing of stimuli and to disordered appraisals of the environment. Problems of social cognition may be newly viewed as disordered embodied communication. The embodiment perspective suggests novel treatment strategies through psychotherapy and body-oriented interventions, and may ultimately provide biomarkers for diagnosis.

A Treatment-Resistant Default Mode Subnetwork in Major Depression

Background Previous studies have suggested that the default mode network (DMN) plays a central role in the physiopathology of major depressive disorder (MDD). However, the effect of antidepressant treatment on functional connectivity within the DMN has yet to be established. Considering the very high rates of relapse in recovered subjects, we hypothesized that abnormalities in DMN functional connectivity would persist in recovered MDD subjects. Methods Resting state functional magnetic resonance imaging images were collected from 24 MDD patients and 29 healthy control subjects. After 12 weeks of antidepressant treatment, 18 recovered MDD subjects were scanned again. Group independent component analysis was performed to decompose the resting state images into spatially independent components. Default mode subnetworks were identified using a template based on previous studies. Group differences in the ensuing subnetworks were tested using two-sample t tests. Results Two spatially independent default mode subnetworks were detected in all subjects: the anterior subnetwork and the posterior subnetwork. Both subnetworks showed increased functional connectivity in pretreatment MDD subjects, relative to control subjects. Differences in the posterior subnetwork were normalized after antidepressant treatment, while abnormal functional connectivity persisted within the anterior subnetwork. Conclusions Our findings suggest a dissociation of the DMN into subnetworks, where persistent abnormal functional connectivity within the anterior subnetwork in recovered MDD subjects may constitute a biomarker of asymptomatic depression and potential for relapse.

Multimodal Image Coregistration and Partitioning—A Unified Framework

This paper presents a method for the coregistration and partitioning (i.e., tissue segmentation) of brain images that have been acquired in different modalities. The basic idea is that instead of matching two images directly, one performs intermediate within-modality registrations to two template images that are already in register. One can use a least-squares minimization to determine the affine transformations that map between the templates and the images. By incorporating suitable constraints, a rigid body transformation which directly maps between the images can be extracted from these more general affine transformations. A further refinement capitalizes on the implicit normalization of both images into a standard space. This facilitates segmentation or partitioning of both original images into homologous tissue classifications. Once partitioned, the partitions can be jointly matched, further increasing the accuracy of the coregistration. In short, these techniques reduce the between-modality problem to a series of simpler within-modality problems. These methods are relatively robust, address a number of problems in image transformations, and require no manual intervention.

The evolution of brain architectures for predictive coding and active inference

This article considers the evolution of brain architectures for predictive processing. We argue that brain mechanisms for predictive perception and action are not late evolutionary additions of advanced creatures like us. Rather, they emerged gradually from simpler predictive loops (e.g. autonomic and motor reflexes) that were a legacy from our earlier evolutionary ancestors—and were key to solving their fundamental problems of adaptive regulation. We characterize simpler-to-more-complex brains formally, in terms of generative models that include predictive loops of increasing hierarchical breadth and depth. These may start from a simple homeostatic motif and be elaborated during evolution in four main ways: these include the multimodal expansion of predictive control into an allostatic loop; its duplication to form multiple sensorimotor loops that expand an animal's behavioural repertoire; and the gradual endowment of generative models with hierarchical depth (to deal with aspects of the world that unfold at different spatial scales) and temporal depth (to select plans in a future-oriented manner). In turn, these elaborations underwrite the solution to biological regulation problems faced by increasingly sophisticated animals. Our proposal aligns neuroscientific theorising—about predictive processing—with evolutionary and comparative data on brain architectures in different animal species. This article is part of the theme issue ‘Systems neuroscience through the lens of evolutionary theory’.

Gradient-Free De Novo Learning

This technical note applies active inference to the problem of learning goal-directed behaviour from scratch, namely, de novo learning. By de novo learning, we mean discovering, directly from observations, the structure and parameters of a discrete generative model for sequential policy optimisation. Concretely, our procedure grows and then reduces a model until it discovers a pullback attractor over (generalised) states; this attracting set supplies paths of least action among goal states while avoiding costly states. The implicit efficiency rests upon reframing the learning problem through the lens of the free energy principle, under which it is sufficient to learn a generative model whose dynamics feature such an attracting set. For context, we briefly relate this perspective to value-based formulations (e.g., Bellman optimality) and then apply the active inference formulation to a small arcade game to illustrate de novo structure learning and ensuing agency.

Bayesian Modelling of Induced Responses and Neuronal Rhythms

Neural rhythms or oscillations are ubiquitous in neuroimaging data. These spectral responses have been linked to several cognitive processes; including working memory, attention, perceptual binding and neuronal coordination. In this paper, we show how Bayesian methods can be used to finesse the ill-posed problem of reconstructing-and explaining-oscillatory responses. We offer an overview of recent developments in this field, focusing on (i) the use of MEG data and Empirical Bayes to build hierarchical models for group analyses-and the identification of important sources of inter-subject variability and (ii) the construction of novel dynamic causal models of intralaminar recordings to explain layer-specific activity. We hope to show that electrophysiological measurements contain much more spatial information than is often thought: on the one hand, the dynamic causal modelling of non-invasive (low spatial resolution) electrophysiology can afford sub-millimetre (hyper-acute) resolution that is limited only by the (spatial) complexity of the underlying (dynamic causal) forward model. On the other hand, invasive microelectrode recordings (that penetrate different cortical layers) can reveal laminar-specific responses and elucidate hierarchical message passing and information processing within and between cortical regions at a macroscopic scale. In short, the careful and biophysically grounded modelling of sparse data enables one to characterise the neuronal architectures generating oscillations in a remarkable detail.

Bayesian Sparsification for Deep Neural Networks With Bayesian Model Reduction

Deep learning's immense capabilities are often constrained by the complexity of its models, leading to an increasing demand for effective sparsification techniques.Bayesian sparsification for deep learning emerges as a crucial approach, facilitating the design of models that are both computationally efficient and competitive in terms of performance across various deep learning applications.The state-of-the-art -in Bayesian sparsification of deep neural networks -combines structural shrinkage priors on model weights with an approximate inference scheme based on stochastic variational inference.However, model inversion of the full generative model is exceptionally computationally demanding, especially when compared to standard deep learning of point estimates.In this context, we advocate for the use of Bayesian model reduction (BMR) as a more efficient alternative for pruning of model weights.As a generalization of the Savage-Dickey ratio, BMR allows a post-hoc elimination of redundant model weights based on the posterior estimates under a straightforward (non-hierarchical) generative model.Our comparative study highlights the advantages of the BMR method relative to established approaches based on hierarchical horseshoe priors over model weights.We illustrate the potential of BMR across various deep learning architectures, from classical networks like LeNet to modern frameworks such as Vision Transformers and MLP-Mixers.

Investigating individual differences in brain abnormalities in autism

Abstract Autism is a psychiatric syndrome characterized by impairments in three domains: social interaction, communication, and restricted and repetitive behaviours and interests. Recent findings implicate the amygdala in the neurobiology of autism. In this paper, we report the results of a series of novel experimental investigations focusing on the structure and function of the amygdala in a group of children with autism. The first section attempts to determine if abnormality of the amygdala can be identified in an individual using magnetic resonance imaging in vivo. Using single-case voxel-based morphometric analyses, abnormality in the amygdala was detected in half the children with autism. Abnormalities in other regions were also found. In the second section, emotional modulation of the startle response was investigated in the group of autistic children. Surprisingly, there were no significant differences between the patterns of emotional modulation of the startle response in the autistic group compared with the controls.

Variational Representational Similarity Analysis (vRSA) for M/EEG

This paper introduces variational representational similarity analysis RSA (vRSA) for electromagnetic recordings of neural responses (e.g., EEG, MEG, ECoG or LFP). Variational RSA is a Bayesian approach for testing whether the similarity of stimuli or experimental conditions is expressed in univariate or multivariate neural recordings. Extending an approach previously introduced in the context of functional MRI, vRSA decomposes the condition-by-condition data covariance matrix into hypothesised effects and observation noise, thereby casting RSA as a covariance component estimation problem. In this context, peristimulus time may be treated as an experimental factor, enabling one to test for the probability that different experimental effects are expressed in data at different times. Variational Bayesian methods are used for model estimation and model comparison, which confer a number of advantages over classical approaches, including statistically efficient hypothesis testing, quantification of uncertainty using Bayesian credible intervals and computational efficiency. After introducing the theory, we provide a worked example using openly available EEG data. Software functions implementing vRSA for the SPM software package accompany this paper, together with exemplar analysis scripts.

Designing Explainable Artificial Intelligence with Active Inference: A Framework for Transparent Introspection and Decision-Making

An Active Inference Approach to Modeling Structure Learning: Concept Learning as an Example Case

Within computational neuroscience, the algorithmic and neural basis of structure learning remains poorly understood. Concept learning is one primary example, which requires both a type of internal model expansion process (adding novel hidden states that explain new observations), and a model reduction process (merging different states into one underlying cause and thus reducing model complexity via meta-learning). Although various algorithmic models of concept learning have been proposed within machine learning and cognitive science, many are limited to various degrees by an inability to generalize, the need for very large amounts of training data, and/or insufficiently established biological plausibility. Using concept learning as an example case, we introduce a novel approach for modeling structure learning-and specifically state-space expansion and reduction-within the active inference framework and its accompanying neural process theory. Our aim is to demonstrate its potential to facilitate a novel line of active inference research in this area. The approach we lay out is based on the idea that a generative model can be equipped with extra (hidden state or cause) "slots" that can be engaged when an agent learns about novel concepts. This can be combined with a Bayesian model reduction process, in which any concept learning-associated with these slots-can be reset in favor of a simpler model with higher model evidence. We use simulations to illustrate this model's ability to add new concepts to its state space (with relatively few observations) and increase the granularity of the concepts it currently possesses. We also simulate the predicted neural basis of these processes. We further show that it can accomplish a simple form of "one-shot" generalization to new stimuli. Although deliberately simple, these simulation results highlight ways in which active inference could offer useful resources in developing neurocomputational models of structure learning. They provide a template for how future active inference research could apply this approach to real-world structure learning problems and assess the added utility it may offer.

Modelling cortical network dynamics

We consider the theoretical constraints on interactions between coupled cortical columns. Each column comprises a set of neural populations, where each population is modelled as a neural mass. The existence of semi-stable states within a cortical column has been shown to be dependent on the type of interaction between the constituent neuronal subpopulations, i.e., the form of the implicit synaptic convolution kernels. Current-to-current coupling has been shown, in contrast to potential-to-current coupling, to create semi-stable states within a cortical column. In this analytic and numerical study, the interaction between semi-stable states is characterized by equations of motion for ensemble activity. We show that for small excitations, the dynamics follow the Kuramoto model. However, in contrast to previous work, we derive coupled equations between phase and amplitude dynamics. This affords the possibility of defining connectivity as a dynamic variable. The turbulent flow of phase dynamics found in networks of Kuramoto oscillators indicate turbulent changes in dynamic connectivity for coupled cortical columns. Crucially, this is something that has been recorded in epileptic seizures. We used the results we derived to estimate a seizure propagation model, which allowed for relatively straightforward inversions using variational Laplace (a.k.a., Dynamic Causal Modelling). The face validity of the ensuing seizure propagation model was established using simulated data as a prelude to future work; which will investigate dynamic connectivity from empirical data. This model also allows predictions of seizure evolution, induced by virtual lesions to synaptic connectivity: something that could be of clinical use, when applied to epilepsy surgical cases.

The Analysis of Brain PET Radioligand Displacement Studies

Supervised structure learning

Tracking slow modulations in synaptic gain using dynamic causal modelling: Validation in epilepsy

In this work we propose a proof of principle that dynamic causal modelling can identify plausible mechanisms at the synaptic level underlying brain state changes over a timescale of seconds. As a benchmark example for validation we used intracranial electroencephalographic signals in a human subject. These data were used to infer the (effective connectivity) architecture of synaptic connections among neural populations assumed to generate seizure activity. Dynamic causal modelling allowed us to quantify empirical changes in spectral activity in terms of a trajectory in parameter space - identifying key synaptic parameters or connections that cause observed signals. Using recordings from three seizures in one patient, we considered a network of two sources (within and just outside the putative ictal zone). Bayesian model selection was used to identify the intrinsic (within-source) and extrinsic (between-source) connectivity. Having established the underlying architecture, we were able to track the evolution of key connectivity parameters (e.g., inhibitory connections to superficial pyramidal cells) and test specific hypotheses about the synaptic mechanisms involved in ictogenesis. Our key finding was that intrinsic synaptic changes were sufficient to explain seizure onset, where these changes showed dissociable time courses over several seconds. Crucially, these changes spoke to an increase in the sensitivity of principal cells to intrinsic inhibitory afferents and a transient loss of excitatory-inhibitory balance.

P.0583 Identifying prodromal biomarkers for schizophrenia and bipolar disorder with magnetoencephalography