Marmaduke Woodman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Marmaduke Woodman has a total of 51 co-authors. Here are the top co-authors by number of shared publications:
- Viktor Jirsa
Aix Marseille University, Institute de Neuroscience des Systems, UMR 1106, INSERM
20 shared publications - Meysam Hashemi
Aix Marseille Univ, INSERM, INS, Inst Neurosci Syst, Marseille, France
15 shared publications - Spase Petkoski
Aix Marseille University, Institute de Neuroscience des Systems, UMR 1106, INSERM
10 shared publications - Abolfazl Ziaeemehr
Aix Marseille University, INSERM, INS, Inst Neurosci System
6 shared publications - Viktor Sip
Aix Marseille Univ, Inserm, INS, Inst Neurosci Syst
5 shared publications - Huifang Wang
Aix Marseille Université, INSERM, INS, Institut de Neurosciences des Systèmes, Marseille, France
4 shared publications
Latest Publications
- Open Access
- 16 September 2026
- Published by Springer Nature in Communications Physics
- Vol. 9 (1) , 304
Data-driven mean-field within whole-brain models
- 16 September 2026
- Published by Springer Nature in Communications Physics
- Vol. 9 (1) , 304
Open AccessData-driven mean-field within whole-brain models
- Open Access
- 27 August 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (124) , 10612
Virtual Brain Inference (VBI) toolkit on EBRAINS
- 27 August 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (124) , 10612
Open AccessVirtual Brain Inference (VBI) toolkit on EBRAINS
- Open Access
Cortico-subcortical multi-head self-attention as a substrate for cognitive performance
The neocortex is central to mammalian cognition, yet a computational framework that is both biologically constrained and capable of performing complex cognitive tasks remains missing. Here we show that cortico-thalamic circuits are well suited to implement multi-head self- and cross-attention, the mechanism underlying the cognitive abilities of transformer networks. We propose that layer 2/3 pyramidal cells maintain a recurrent key-value memory, while layer 5 pyramidal cells decode the memory retrieved by an incoming query. The computation of keys, values and queries maps onto core and matrix thalamo-cortical projections, distributed across the micro- and macro-columns of a cortical area. One cortical area forms an attention head, and cortex a multi-head self-attention network. The same thalamo-cortical microcircuit also calculates sensory prediction errors guiding gradient-based synaptic plasticity. A reward-prediction error gates via basal ganglia the cortical output and the re-activation of hippocampal memories. The trained network aligns with human intracranial recordings during speech perception. Overall, the suggested cortico-subcortical attention circuit may represent a substrate for the cognitive capacity of mammals.
Open AccessCortico-subcortical multi-head self-attention as a substrate for cognitive performance
The neocortex is central to mammalian cognition, yet a computational framework that is both biologically constrained and capable of performing complex cognitive tasks remains missing. Here we show that cortico-thalamic circuits are well suited to implement multi-head self- and cross-attention, the mechanism underlying the cognitive abilities of transformer networks. We propose that layer 2/3 pyramidal cells maintain a recurrent key-value memory, while layer 5 pyramidal cells decode the memory retrieved by an incoming query. The computation of keys, values and queries maps onto core and matrix thalamo-cortical projections, distributed across the micro- and macro-columns of a cortical area. One cortical area forms an attention head, and cortex a multi-head self-attention network. The same thalamo-cortical microcircuit also calculates sensory prediction errors guiding gradient-based synaptic plasticity. A reward-prediction error gates via basal ganglia the cortical output and the re-activation of hippocampal memories. The trained network aligns with human intracranial recordings during speech perception. Overall, the suggested cortico-subcortical attention circuit may represent a substrate for the cognitive capacity of mammals.
- Open Access
- 12 December 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Virtual Brain Inference (VBI), a flexible and integrative toolkit for efficient probabilistic inference on whole-brain models
Network neuroscience has proven essential for understanding the principles and mechanisms underlying complex brain (dys)function and cognition. In this context, whole-brain network modeling—also known as virtual brain modeling—combines computational models of brain dynamics (placed at each network node) with individual brain imaging data (to coordinate and connect the nodes), advancing our understanding of the complex dynamics of the brain and its neurobiological underpinnings. However, there remains a critical need for automated model inversion tools to estimate control (bifurcation) parameters at large scales associated with neuroimaging modalities, given their varying spatio-temporal resolutions. This study aims to address this gap by introducing a flexible and integrative toolkit for efficient Bayesian inference on virtual brain models, called Virtual Brain Inference (VBI). This open-source toolkit provides fast simulations, taxonomy of feature extraction, efficient data storage and loading, and probabilistic machine learning algorithms, enabling biophysically interpretable inference from non-invasive and invasive recordings. Through in-silico testing, we demonstrate the accuracy and reliability of inference for commonly used whole-brain network models and their associated neuroimaging data. VBI shows potential to improve hypothesis evaluation in network neuroscience through uncertainty quantification and contribute to advances in precision medicine by enhancing the predictive power of virtual brain models.
- 12 December 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Open AccessVirtual Brain Inference (VBI), a flexible and integrative toolkit for efficient probabilistic inference on whole-brain models
Network neuroscience has proven essential for understanding the principles and mechanisms underlying complex brain (dys)function and cognition. In this context, whole-brain network modeling—also known as virtual brain modeling—combines computational models of brain dynamics (placed at each network node) with individual brain imaging data (to coordinate and connect the nodes), advancing our understanding of the complex dynamics of the brain and its neurobiological underpinnings. However, there remains a critical need for automated model inversion tools to estimate control (bifurcation) parameters at large scales associated with neuroimaging modalities, given their varying spatio-temporal resolutions. This study aims to address this gap by introducing a flexible and integrative toolkit for efficient Bayesian inference on virtual brain models, called Virtual Brain Inference (VBI). This open-source toolkit provides fast simulations, taxonomy of feature extraction, efficient data storage and loading, and probabilistic machine learning algorithms, enabling biophysically interpretable inference from non-invasive and invasive recordings. Through in-silico testing, we demonstrate the accuracy and reliability of inference for commonly used whole-brain network models and their associated neuroimaging data. VBI shows potential to improve hypothesis evaluation in network neuroscience through uncertainty quantification and contribute to advances in precision medicine by enhancing the predictive power of virtual brain models.
- Open Access
A deformable attractor manifold organizes human resting-state brain dynamics
Intrinsic brain activity is often described as wandering within a continuous multivariate space, yet the organizing principles that constrain these dynamics remain unclear. Here, we show that spontaneous human brain activity during rest is structured by a deformable attractor manifold. Using large-scale fMRI datasets and a latent dynamical model, we find that cortical activity occupies two reproducible regimes: a low-coherence state with a unimodal latent distribution and a high-coherence state that exhibits bimodality, consistent with transient bistability across association networks. A compact two-parameter energy landscape explains these dynamics, revealing that transitions arise not from switching between discrete states, but from continuous deformation of the manifold that reshapes attractor geometry. Excursions into the bistable regime occur as rapid “jumps”, whereas returns follow slow drifts along the manifold, reflecting network-specific timescales. Individuals with greater expression of the bistable regime show higher cognitive fluidity, and manifold parameters differentiate mild cognitive impairment from matched controls. These findings identify an organizing geometric and dynamical principle of resting activity, linking large-scale cortical coordination, cognitive variability, and vulnerability to pathology.
Open AccessA deformable attractor manifold organizes human resting-state brain dynamics
Intrinsic brain activity is often described as wandering within a continuous multivariate space, yet the organizing principles that constrain these dynamics remain unclear. Here, we show that spontaneous human brain activity during rest is structured by a deformable attractor manifold. Using large-scale fMRI datasets and a latent dynamical model, we find that cortical activity occupies two reproducible regimes: a low-coherence state with a unimodal latent distribution and a high-coherence state that exhibits bimodality, consistent with transient bistability across association networks. A compact two-parameter energy landscape explains these dynamics, revealing that transitions arise not from switching between discrete states, but from continuous deformation of the manifold that reshapes attractor geometry. Excursions into the bistable regime occur as rapid “jumps”, whereas returns follow slow drifts along the manifold, reflecting network-specific timescales. Individuals with greater expression of the bistable regime show higher cognitive fluidity, and manifold parameters differentiate mild cognitive impairment from matched controls. These findings identify an organizing geometric and dynamical principle of resting activity, linking large-scale cortical coordination, cognitive variability, and vulnerability to pathology.
- Open Access
- 16 October 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 14
Virtual Brain Inference (VBI): A flexible and integrative toolkit for efficient probabilistic inference on virtual brain models
Network neuroscience has proven essential for understanding the principles and mechanisms underlying complex brain (dys)function and cognition. In this context, whole-brain network modeling–also known as virtual brain modeling–combines computational models of brain dynamics (placed at each network node) with individual brain imaging data (to coordinate and connect the nodes), advancing our understanding of the complex dynamics of the brain and its neurobiological underpinnings. However, there remains a critical need for automated model inversion tools to estimate control (bifurcation) parameters at large scales associated with neuroimaging modalities, given their varying spatio-temporal resolutions. This study aims to address this gap by introducing a flexible and integrative toolkit for efficient Bayesian inference on virtual brain models, called Virtual Brain Inference (VBI). This open-source toolkit provides fast simulations, taxonomy of feature extraction, efficient data storage and loading, and probabilistic machine learning algorithms, enabling biophysically interpretable inference from non-invasive and invasive recordings. Through in-silico testing, we demonstrate the accuracy and reliability of inference for commonly used whole-brain network models and their associated neuroimaging data. VBI shows potential to improve hypothesis evaluation in network neuroscience through uncertainty quantification, and contribute to advances in precision medicine by enhancing the predictive power of virtual brain models.
- 16 October 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 14
Open AccessVirtual Brain Inference (VBI): A flexible and integrative toolkit for efficient probabilistic inference on virtual brain models
Network neuroscience has proven essential for understanding the principles and mechanisms underlying complex brain (dys)function and cognition. In this context, whole-brain network modeling–also known as virtual brain modeling–combines computational models of brain dynamics (placed at each network node) with individual brain imaging data (to coordinate and connect the nodes), advancing our understanding of the complex dynamics of the brain and its neurobiological underpinnings. However, there remains a critical need for automated model inversion tools to estimate control (bifurcation) parameters at large scales associated with neuroimaging modalities, given their varying spatio-temporal resolutions. This study aims to address this gap by introducing a flexible and integrative toolkit for efficient Bayesian inference on virtual brain models, called Virtual Brain Inference (VBI). This open-source toolkit provides fast simulations, taxonomy of feature extraction, efficient data storage and loading, and probabilistic machine learning algorithms, enabling biophysically interpretable inference from non-invasive and invasive recordings. Through in-silico testing, we demonstrate the accuracy and reliability of inference for commonly used whole-brain network models and their associated neuroimaging data. VBI shows potential to improve hypothesis evaluation in network neuroscience through uncertainty quantification, and contribute to advances in precision medicine by enhancing the predictive power of virtual brain models.
- 12 September 2025
- Published by Springer Nature
- p. 78-88
Amortizing Personalization in Virtual Brain Twins
- 12 September 2025
- Published by Springer Nature
- p. 78-88
Amortizing Personalization in Virtual Brain Twins
- Open Access
Amortizing personalization in virtual brain twins
Virtual brain twins are personalized digital models of individual human subject or patient's brains, allowing for mechanistic interpretation of neuroimaging data features. Training and inference with these models however presents a pair of challenges: large shared infrastructure do not allow for use of personal data and inference in clinical applications should not require significant resources. We introduce "anonymized personalization" to address both by expanding model priors to include personalization which under amortized inference allows training to be performed anonymously, while inference is both personalized and lightweight. We illustrate the basic approach, demonstrate reliability in an example, and discuss the impact on both experimental and computational neuroscience. Code is available at https://github.com/ins-amu/apvbt.
Open AccessAmortizing personalization in virtual brain twins
Virtual brain twins are personalized digital models of individual human subject or patient's brains, allowing for mechanistic interpretation of neuroimaging data features. Training and inference with these models however presents a pair of challenges: large shared infrastructure do not allow for use of personal data and inference in clinical applications should not require significant resources. We introduce "anonymized personalization" to address both by expanding model priors to include personalization which under amortized inference allows training to be performed anonymously, while inference is both personalized and lightweight. We illustrate the basic approach, demonstrate reliability in an example, and discuss the impact on both experimental and computational neuroscience. Code is available at https://github.com/ins-amu/apvbt.
- Open Access
- 4 June 2025
- Published by The Royal Society in Journal of The Royal Society Interface
- Vol. 22 (227) , 20240880
Dynamic causal modelling in probabilistic programming languages
Understanding the intricate dynamics of brain activities necessitates models that incorporate causality and nonlinearity. Dynamic causal modelling (DCM) presents a statistical framework that embraces causal relationships among brain regions and their responses to experimental manipulations, such as stimulation. In this study, we perform Bayesian inference on a neurobiologically plausible generative model that simulates event-related potentials observed in magneto/encephalography data. This translates into probabilistic inference of latent and observed states of a system driven by input stimuli, described by a set of nonlinear ordinary differential equations (ODEs) and potentially correlated parameters. We provide a guideline for reliable inference in the presence of multimodality, which arises from parameter degeneracy, ultimately enhancing the predictive accuracy of neural dynamics. Solutions include optimizing the hyperparameters, leveraging initialization with prior information and employing weighted stacking based on predictive accuracy. Moreover, we implement the inference and conduct comprehensive model comparison in several probabilistic programming languages to streamline the process and benchmark their efficiency. Our investigation shows that model inversion in DCM extends beyond variational approximation frameworks, demonstrating the effectiveness of gradient-based Markov chain Monte Carlo methods. We illustrate the accuracy and efficiency of posterior estimation using a self-tuning variant of Hamiltonian Monte Carlo and the automatic Laplace approximation, effectively addressing parameter degeneracy challenges. This technical endeavour holds the potential to advance the inversion of state-space ODE models, and contribute to neuroscience research and applications in neuroimaging through automatic DCM.
- 4 June 2025
- Published by The Royal Society in Journal of The Royal Society Interface
- Vol. 22 (227) , 20240880
Open AccessDynamic causal modelling in probabilistic programming languages
Understanding the intricate dynamics of brain activities necessitates models that incorporate causality and nonlinearity. Dynamic causal modelling (DCM) presents a statistical framework that embraces causal relationships among brain regions and their responses to experimental manipulations, such as stimulation. In this study, we perform Bayesian inference on a neurobiologically plausible generative model that simulates event-related potentials observed in magneto/encephalography data. This translates into probabilistic inference of latent and observed states of a system driven by input stimuli, described by a set of nonlinear ordinary differential equations (ODEs) and potentially correlated parameters. We provide a guideline for reliable inference in the presence of multimodality, which arises from parameter degeneracy, ultimately enhancing the predictive accuracy of neural dynamics. Solutions include optimizing the hyperparameters, leveraging initialization with prior information and employing weighted stacking based on predictive accuracy. Moreover, we implement the inference and conduct comprehensive model comparison in several probabilistic programming languages to streamline the process and benchmark their efficiency. Our investigation shows that model inversion in DCM extends beyond variational approximation frameworks, demonstrating the effectiveness of gradient-based Markov chain Monte Carlo methods. We illustrate the accuracy and efficiency of posterior estimation using a self-tuning variant of Hamiltonian Monte Carlo and the automatic Laplace approximation, effectively addressing parameter degeneracy challenges. This technical endeavour holds the potential to advance the inversion of state-space ODE models, and contribute to neuroscience research and applications in neuroimaging through automatic DCM.
- Top Cited PapersOpen Access
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Reviews in Biomedical Engineering
- Vol. 19 (19373333) , 111-139
Principles and Operation of Virtual Brain Twins
Current clinical methods often overlook individual variability by relying on population-wide trials, while mechanismbased trials remain underutilized in neuroscience due to the brain's complexity. This situation may change through the use of a Virtual Brain Twin (VBT), which is a personalized digital replica of an individual's brain, integrating structural and functional brain data into advanced computational models and inference algorithms. By bridging the gap between molecular mechanisms, whole-brain dynamics, and imaging data, VBTs enhance the understanding of (patho)physiological mechanisms, advancing insights into both healthy and disordered brain function. Central to VBT is the network modeling that couples mesoscopic representation of neuronal activity through white matter connectivity, enabling the simulation of brain dynamics at a network level. This transformative approach provides interpretable predictive capabilities, supporting clinicians in personalizing treatments and optimizing interventions. This Review outlines the key components of VBT development, covering the conceptual, mathematical, technical, and clinical aspects. We describe the stages of VBT construction-from anatomical coupling and modeling to simulation and Bayesian inference-and demonstrate their applications in resting-state, healthy aging, multiple sclerosis, and epilepsy. Finally, we discuss potential extensions to other neurological disorders, such as Parkinson's disease, and explore future applications in consciousness research and brain-computer interfaces, paving the way for advancements in personalized medicine and brainmachine integration.
- 21 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Reviews in Biomedical Engineering
- Vol. 19 (19373333) , 111-139
Top Cited PapersOpen AccessPrinciples and Operation of Virtual Brain Twins
Current clinical methods often overlook individual variability by relying on population-wide trials, while mechanismbased trials remain underutilized in neuroscience due to the brain's complexity. This situation may change through the use of a Virtual Brain Twin (VBT), which is a personalized digital replica of an individual's brain, integrating structural and functional brain data into advanced computational models and inference algorithms. By bridging the gap between molecular mechanisms, whole-brain dynamics, and imaging data, VBTs enhance the understanding of (patho)physiological mechanisms, advancing insights into both healthy and disordered brain function. Central to VBT is the network modeling that couples mesoscopic representation of neuronal activity through white matter connectivity, enabling the simulation of brain dynamics at a network level. This transformative approach provides interpretable predictive capabilities, supporting clinicians in personalizing treatments and optimizing interventions. This Review outlines the key components of VBT development, covering the conceptual, mathematical, technical, and clinical aspects. We describe the stages of VBT construction-from anatomical coupling and modeling to simulation and Bayesian inference-and demonstrate their applications in resting-state, healthy aging, multiple sclerosis, and epilepsy. Finally, we discuss potential extensions to other neurological disorders, such as Parkinson's disease, and explore future applications in consciousness research and brain-computer interfaces, paving the way for advancements in personalized medicine and brainmachine integration.