Jean-Philippe Thivierge
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jean-Philippe Thivierge has a total of 62 co-authors. Here are the top co-authors by number of shared publications:
Camille Godin
University of Ottawa
6 shared publicationsArtem Pilzak
School of Psychology, University of Ottawa, Ottawa, Canada
5 shared publications- Eric S. Kuebler
Robarts Research Institute, London, Ontario, Canada
5 shared publications Matias Calderini
School of Psychology, University of Ottawa 1 , 156 Jean-Jacques Lussier, Ottawa, Ontario K1N 6N5
4 shared publications- Jean-Claude Béïque
Department of Cellular and Molecular Medicine, University of Ottawa
3 shared publications Joseph S. Tauskela
Department of Translational Bioscience, Human Health Therapeutics, National Research Council of Canada, Ottawa, Ontario, Canada
3 shared publications
Latest Publications
- Open Access
CLEAR-FORCE: A reservoir computing framework for closed-loop brain stimulation
Electrical stimulation of brain activity has emerged as a key therapeutic intervention for neurological diseases. The success of these approaches relies on precisely tuned stimulation parameters. A major challenge is that these parameters should ideally be adjusted in real-time based on changes in brain activity. Closed-loop controllers address this challenge by monitoring brain activity and adapting stimulation according to ongoing changes in neural dynamics. While machine learning has been helpful in this context, closed-loop controllers typically do not adapt in real-time to fluctuations in brain activity. Here, we propose a novel framework based on reservoir computing to train a closed-loop, real-time neural controller. This approach, termed CLEAR-FORCE, trains the output weights of a recurrent network based on a combination of target signals and feedback from brain activity. Two series of simulations were performed, where the controller was successful at (1) inducing neural oscillations in a population of spiking neurons and (2) suppressing pathological oscillations. This framework provides a promising avenue for the implementation of neural controllers in low-powered, real-time, and adaptive hardware that harnesses machine learning for the personalized treatment of brain disorders.
Open AccessCLEAR-FORCE: A reservoir computing framework for closed-loop brain stimulation
Electrical stimulation of brain activity has emerged as a key therapeutic intervention for neurological diseases. The success of these approaches relies on precisely tuned stimulation parameters. A major challenge is that these parameters should ideally be adjusted in real-time based on changes in brain activity. Closed-loop controllers address this challenge by monitoring brain activity and adapting stimulation according to ongoing changes in neural dynamics. While machine learning has been helpful in this context, closed-loop controllers typically do not adapt in real-time to fluctuations in brain activity. Here, we propose a novel framework based on reservoir computing to train a closed-loop, real-time neural controller. This approach, termed CLEAR-FORCE, trains the output weights of a recurrent network based on a combination of target signals and feedback from brain activity. Two series of simulations were performed, where the controller was successful at (1) inducing neural oscillations in a population of spiking neurons and (2) suppressing pathological oscillations. This framework provides a promising avenue for the implementation of neural controllers in low-powered, real-time, and adaptive hardware that harnesses machine learning for the personalized treatment of brain disorders.
- Open Access
- 4 October 2026
- Published by Springer Nature in Scientific Reports
Scale-free dynamics and reservoir computing in a non-stationary Freedericksz model
- 4 October 2026
- Published by Springer Nature in Scientific Reports
Open AccessScale-free dynamics and reservoir computing in a non-stationary Freedericksz model
- Open Access
- 1 September 2026
- Published by IOP Publishing in Journal of Physics: Complexity
- Vol. 7 (3) , 035014
A stochastic Barkley model for noise-driven intermittent waves
Spiral waves are a prominent form of spatiotemporal organization in excitable media. In brain networks, these waves are characterized by intermittent events separated by quiescent periods, which are not captured by models that generate sustained spiral activity. Here, we develop a stochastic reaction-diffusion model based on the Barkley framework to examine how transient waves emerge from a stable resting state. The model includes an activator variable, a recovery variable, a slow adaptation variable, and stochastic forcing. Numerical simulations reveal three dynamical regimes: a quiescent regime, a regime with sustained waves, and an intermittent regime where transient events emerge and terminate. In the intermittent regime, a subset of wave events exhibit broad phase coverage, rotational structure, and sustained phase winding. Linear stability analysis shows that the quiescent state is asymptotically stable, indicating that wave initiation does not arise from a local instability but from noise-driven threshold crossings. A first-passage approximation predicts an exponential dependence of initiation rates on noise amplitude, damping, and excitation threshold, in agreement with numerical simulations. Together, these results identify a novel stochastic threshold-crossing regime that combines stable quiescence, noise-driven wave nucleation, and slow adaptive recovery to generate intermittent wave dynamics, including events with transient spiral organization.
- 1 September 2026
- Published by IOP Publishing in Journal of Physics: Complexity
- Vol. 7 (3) , 035014
Open AccessA stochastic Barkley model for noise-driven intermittent waves
Spiral waves are a prominent form of spatiotemporal organization in excitable media. In brain networks, these waves are characterized by intermittent events separated by quiescent periods, which are not captured by models that generate sustained spiral activity. Here, we develop a stochastic reaction-diffusion model based on the Barkley framework to examine how transient waves emerge from a stable resting state. The model includes an activator variable, a recovery variable, a slow adaptation variable, and stochastic forcing. Numerical simulations reveal three dynamical regimes: a quiescent regime, a regime with sustained waves, and an intermittent regime where transient events emerge and terminate. In the intermittent regime, a subset of wave events exhibit broad phase coverage, rotational structure, and sustained phase winding. Linear stability analysis shows that the quiescent state is asymptotically stable, indicating that wave initiation does not arise from a local instability but from noise-driven threshold crossings. A first-passage approximation predicts an exponential dependence of initiation rates on noise amplitude, damping, and excitation threshold, in agreement with numerical simulations. Together, these results identify a novel stochastic threshold-crossing regime that combines stable quiescence, noise-driven wave nucleation, and slow adaptive recovery to generate intermittent wave dynamics, including events with transient spiral organization.
- Open Access
- 25 August 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (8) , e1014719
Manifold-constrained plasticity enables stable learning in recurrent neural circuits
The activity of large neuronal populations is often confined to low-dimensional manifolds that can drift over time, posing a challenge for learning rules that assume stable, full-rank representations. Here, we introduce SPLiT (Synaptic Projection Learning with intrinsic Tracking), a synaptic plasticity rule for recurrent neural networks that combines unsupervised manifold tracking with supervised learning. SPLiT uses an online Oja rule to continuously estimate the intrinsic low-dimensional activity subspace and performs normalized least-mean-squares learning in manifold coordinates. In this way, SPLiT keeps synaptic weights aligned with evolving population dynamics. Using a rate-based recurrent network, we show that SPLiT reliably learns time-varying target signals under both constrained dynamics, where activity is restricted to a fixed low-dimensional manifold, and unconstrained dynamics exhibiting changes in the dominant activity subspace. We show that manifold-constrained learning with SPLiT yields faster convergence, less sensitivity to recurrent gain, smaller weight updates, and is more robust to noisy teaching signals. Analytical results show that SPLiT learns the optimal decoder projected onto the instantaneous principal subspace and maintains bounded error under drift. Together, these findings provide a mechanistic account of how synaptic plasticity can leverage the structure of neural manifolds to enable stable and efficient supervised learning despite representational drift.
- 25 August 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (8) , e1014719
Open AccessManifold-constrained plasticity enables stable learning in recurrent neural circuits
The activity of large neuronal populations is often confined to low-dimensional manifolds that can drift over time, posing a challenge for learning rules that assume stable, full-rank representations. Here, we introduce SPLiT (Synaptic Projection Learning with intrinsic Tracking), a synaptic plasticity rule for recurrent neural networks that combines unsupervised manifold tracking with supervised learning. SPLiT uses an online Oja rule to continuously estimate the intrinsic low-dimensional activity subspace and performs normalized least-mean-squares learning in manifold coordinates. In this way, SPLiT keeps synaptic weights aligned with evolving population dynamics. Using a rate-based recurrent network, we show that SPLiT reliably learns time-varying target signals under both constrained dynamics, where activity is restricted to a fixed low-dimensional manifold, and unconstrained dynamics exhibiting changes in the dominant activity subspace. We show that manifold-constrained learning with SPLiT yields faster convergence, less sensitivity to recurrent gain, smaller weight updates, and is more robust to noisy teaching signals. Analytical results show that SPLiT learns the optimal decoder projected onto the instantaneous principal subspace and maintains bounded error under drift. Together, these findings provide a mechanistic account of how synaptic plasticity can leverage the structure of neural manifolds to enable stable and efficient supervised learning despite representational drift.
- Open Access
- 19 March 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Intrinsic stabilization of synaptic plasticity improves learning and robustness in artificial neural networks
Synaptic plasticity, defined as the ability of connections between neurons to strengthen or weaken over time, is essential for learning, yet unregulated plasticity can lead to instability. Inspired by biological feedback mechanisms, we propose a framework that stabilizes synaptic changes in artificial neural networks through top-down signals derived from the network’s own output. The model, termed intrinsic Top-Down Stabilization (iTDS), augments traditional supervised learning by introducing a slower, top-down signal that tracks the network’s output over time and modulates synaptic updates. Instead of directly influencing sensory inputs, these top-down signals guide synaptic plasticity by gradually aligning with ongoing network activity. We demonstrate that this simple mechanism improves training efficiency, enhances generalization, and increases resilience to noise perturbations. The approach is evaluated across recurrent, feedforward, and reservoir networks on 16 synthetic and real-world tasks, including temporal signal encoding, synthetic data categorization, handwritten character recognition, and image classification. Our results clarify how network activity influences the alignment of top-down signals with supervised learning, providing testable predictions for how feedback projections can stabilize learning in biological networks.
- 19 March 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessIntrinsic stabilization of synaptic plasticity improves learning and robustness in artificial neural networks
Synaptic plasticity, defined as the ability of connections between neurons to strengthen or weaken over time, is essential for learning, yet unregulated plasticity can lead to instability. Inspired by biological feedback mechanisms, we propose a framework that stabilizes synaptic changes in artificial neural networks through top-down signals derived from the network’s own output. The model, termed intrinsic Top-Down Stabilization (iTDS), augments traditional supervised learning by introducing a slower, top-down signal that tracks the network’s output over time and modulates synaptic updates. Instead of directly influencing sensory inputs, these top-down signals guide synaptic plasticity by gradually aligning with ongoing network activity. We demonstrate that this simple mechanism improves training efficiency, enhances generalization, and increases resilience to noise perturbations. The approach is evaluated across recurrent, feedforward, and reservoir networks on 16 synthetic and real-world tasks, including temporal signal encoding, synthetic data categorization, handwritten character recognition, and image classification. Our results clarify how network activity influences the alignment of top-down signals with supervised learning, providing testable predictions for how feedback projections can stabilize learning in biological networks.
- 20 January 2026
- Published by Springer Nature in Journal of Biological Physics
- Vol. 52 (1) , 4
Phase gradient modulation of spiral waves in cortical circuits using the complex Ginzburg–Landau equation
Disinhibited brain networks exhibit various forms of spatiotemporal waves, including complex spiral waves that evolve around a fixed spatial locus. In experiments, spiral waves are observed to alter their direction of rotation over time, for instance producing a series of waves with clockwise cycles before switching to waves rotating counterclockwise, or vice-versa. To capture this effect, we developed a model based on the Complex Ginzburg–Landau equation (CGLE). By introducing a modulation in the phase gradient of the complex field that reflects global, time-dependent fluctuations in the surrounding environment, the model produced waves that alternated in their direction of rotation. The rate of alternations was directly proportional to the amplitude of phase modulation. Conditions were explored for the emergence of quasi-stationary frozen waves and noise-induced quenching of spiral waves. Overall, the modified CGLE model provides a candidate mechanism for the emergence of spiral waves where rotational directions are dynamically altered, yielding rich forms of activity that account for the spatiotemporal patterns observed in disinhibited brain circuits.
- 20 January 2026
- Published by Springer Nature in Journal of Biological Physics
- Vol. 52 (1) , 4
Phase gradient modulation of spiral waves in cortical circuits using the complex Ginzburg–Landau equation
Disinhibited brain networks exhibit various forms of spatiotemporal waves, including complex spiral waves that evolve around a fixed spatial locus. In experiments, spiral waves are observed to alter their direction of rotation over time, for instance producing a series of waves with clockwise cycles before switching to waves rotating counterclockwise, or vice-versa. To capture this effect, we developed a model based on the Complex Ginzburg–Landau equation (CGLE). By introducing a modulation in the phase gradient of the complex field that reflects global, time-dependent fluctuations in the surrounding environment, the model produced waves that alternated in their direction of rotation. The rate of alternations was directly proportional to the amplitude of phase modulation. Conditions were explored for the emergence of quasi-stationary frozen waves and noise-induced quenching of spiral waves. Overall, the modified CGLE model provides a candidate mechanism for the emergence of spiral waves where rotational directions are dynamically altered, yielding rich forms of activity that account for the spatiotemporal patterns observed in disinhibited brain circuits.
- Open Access
- 28 August 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Unsupervised pipeline for the identification of cortical excitatory and inhibitory neurons in high-density multielectrode arrays with ground-truth validation
Large-scale extracellular recording techniques have advanced the study of neuronal circuits but lack methods to reliably identify cell types while scaling to thousands of neurons. We introduce spikeMAP, a pipeline for analyzing large-scale in vitro cortical recordings that combines spike sorting with cell-type identification using viral and optogenetic validation. SpikeMAP integrates data analysis with optogenetic, viral, and pharmacological protocols to dynamically probe distinct cell types while recording from large populations. The pipeline fits spike waveforms using spline interpolation to measure half-amplitude and peak-to-peak durations, applies principal component analysis and k-means clustering to isolate single-neuron signals, and uses linear discriminant analysis to optimize cluster separability. Channel source locations are determined through spatiotemporal spike waveform characteristics. Applied to mouse prefrontal cortex slices recorded on a 4096-channel array, spikeMAP effectively distinguishes regular-spiking excitatory neurons from fast-spiking inhibitory interneurons via action potential waveform, Fano factor, and spatial cross-correlations. This validated toolbox enables comprehensive characterization of neuronal activity across cell types in high-density recordings, offering a scalable approach to study microcircuit interactions in the brain.
- 28 August 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Open AccessUnsupervised pipeline for the identification of cortical excitatory and inhibitory neurons in high-density multielectrode arrays with ground-truth validation
Large-scale extracellular recording techniques have advanced the study of neuronal circuits but lack methods to reliably identify cell types while scaling to thousands of neurons. We introduce spikeMAP, a pipeline for analyzing large-scale in vitro cortical recordings that combines spike sorting with cell-type identification using viral and optogenetic validation. SpikeMAP integrates data analysis with optogenetic, viral, and pharmacological protocols to dynamically probe distinct cell types while recording from large populations. The pipeline fits spike waveforms using spline interpolation to measure half-amplitude and peak-to-peak durations, applies principal component analysis and k-means clustering to isolate single-neuron signals, and uses linear discriminant analysis to optimize cluster separability. Channel source locations are determined through spatiotemporal spike waveform characteristics. Applied to mouse prefrontal cortex slices recorded on a 4096-channel array, spikeMAP effectively distinguishes regular-spiking excitatory neurons from fast-spiking inhibitory interneurons via action potential waveform, Fano factor, and spatial cross-correlations. This validated toolbox enables comprehensive characterization of neuronal activity across cell types in high-density recordings, offering a scalable approach to study microcircuit interactions in the brain.
- Open Access
- 13 August 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
SpikeMAP: An unsupervised pipeline for the identification of cortical excitatory and inhibitory neurons in high-density multielectrode arrays with ground-truth validation
Large-scale extracellular recording techniques represent a major advance in interrogating the structure and dynamics of neuronal circuits. However, methods that can resolve cell-type identity in a principled way, while simultaneously scaling to thousands of neurons, are currently lacking. Here, we introduce spikeMAP, a pipeline for the analysis of large-scale recordings of in vitro cortical activity that not only allows for the detection of spikes produced by single neurons (spike sorting), but also allows for the reliable distinction between genetically determined cell types by utilizing viral and optogenetic strategies as ground-truth validation. This approach tightly integrates the data analysis pipeline to an optogenetic, viral, and pharmacological protocol allowing for the dynamical probing of distinct cell-types while simultaneously recording from large populations. The novelty of spikeMAP is to combine a stream of well-established analysis techniques in an end-to-end fashion, creating a unified framework as follows. First, individual spike waveforms are fitted by spline interpolation to estimate their half-amplitude and peak-to-peak durations. These values are then entered in a principal component analysis with k-means clustering to identify uncorrelated signals from single channels on the array. Optimal separability of clusters is assessed by linear discriminant analysis. Finally, each channel’s source location is identified using spatiotemporal characteristics of spike waveforms across the array. We show that spikeMAP can resolve cell type identity in high-density arrays by analyzing activity monitored from mouse prefrontal cortex in vitro slices with an array of 4,096 closely-spaced channels. Using an optotagging functional strategy, we show an effective distinction of regular-spiking excitatory neurons from fast-spiking inhibitory interneurons using measures of action potential waveform, Fano factor, and spatially-dependent cross-correlations. In sum, the approach introduces a toolbox, validated by an experimental pipeline, that allows for a comprehensive characterization of neuronal activity obtained from different cell-types in high-density multielectrode recordings. This provides a scalable approach to investigate the interplay between distinct cell types in microcircuits of the brain.
- 13 August 2025
- Published by eLife Sciences Publications, Ltd in eLife preprints
Open AccessSpikeMAP: An unsupervised pipeline for the identification of cortical excitatory and inhibitory neurons in high-density multielectrode arrays with ground-truth validation
Large-scale extracellular recording techniques represent a major advance in interrogating the structure and dynamics of neuronal circuits. However, methods that can resolve cell-type identity in a principled way, while simultaneously scaling to thousands of neurons, are currently lacking. Here, we introduce spikeMAP, a pipeline for the analysis of large-scale recordings of in vitro cortical activity that not only allows for the detection of spikes produced by single neurons (spike sorting), but also allows for the reliable distinction between genetically determined cell types by utilizing viral and optogenetic strategies as ground-truth validation. This approach tightly integrates the data analysis pipeline to an optogenetic, viral, and pharmacological protocol allowing for the dynamical probing of distinct cell-types while simultaneously recording from large populations. The novelty of spikeMAP is to combine a stream of well-established analysis techniques in an end-to-end fashion, creating a unified framework as follows. First, individual spike waveforms are fitted by spline interpolation to estimate their half-amplitude and peak-to-peak durations. These values are then entered in a principal component analysis with k-means clustering to identify uncorrelated signals from single channels on the array. Optimal separability of clusters is assessed by linear discriminant analysis. Finally, each channel’s source location is identified using spatiotemporal characteristics of spike waveforms across the array. We show that spikeMAP can resolve cell type identity in high-density arrays by analyzing activity monitored from mouse prefrontal cortex in vitro slices with an array of 4,096 closely-spaced channels. Using an optotagging functional strategy, we show an effective distinction of regular-spiking excitatory neurons from fast-spiking inhibitory interneurons using measures of action potential waveform, Fano factor, and spatially-dependent cross-correlations. In sum, the approach introduces a toolbox, validated by an experimental pipeline, that allows for a comprehensive characterization of neuronal activity obtained from different cell-types in high-density multielectrode recordings. This provides a scalable approach to investigate the interplay between distinct cell types in microcircuits of the brain.
- Open Access
- 13 August 2025
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 19, 1639829
Maximum likelihood estimation of spatially dependent interactions in large populations of cortical neurons
Understanding how functional connectivity between cortical neurons varies with spatial distance is crucial for characterizing large-scale neural dynamics. However, inferring these spatial patterns is challenging when spike trains are collected from large populations of neurons. Here, we present a maximum likelihood estimation (MLE) framework to quantify distance-dependent functional interactions directly from observed spiking activity. We validate this method using both synthetic spike trains generated from a linear Poisson model and biologically realistic simulations performed with Izhikevich neurons. We then apply the approach to large-scale electrophysiological recordings from V1 cortical neurons. Our results show that the proposed MLE approach robustly captures spatial decay in functional connectivity, providing insights into the spatial structure of population-level neural interactions.
- 13 August 2025
- Published by Frontiers Media SA in Frontiers in Computational Neuroscience
- Vol. 19, 1639829
Open AccessMaximum likelihood estimation of spatially dependent interactions in large populations of cortical neurons
Understanding how functional connectivity between cortical neurons varies with spatial distance is crucial for characterizing large-scale neural dynamics. However, inferring these spatial patterns is challenging when spike trains are collected from large populations of neurons. Here, we present a maximum likelihood estimation (MLE) framework to quantify distance-dependent functional interactions directly from observed spiking activity. We validate this method using both synthetic spike trains generated from a linear Poisson model and biologically realistic simulations performed with Izhikevich neurons. We then apply the approach to large-scale electrophysiological recordings from V1 cortical neurons. Our results show that the proposed MLE approach robustly captures spatial decay in functional connectivity, providing insights into the spatial structure of population-level neural interactions.
- 21 July 2025
- Published by Springer Nature
- p. 258-273
Top-Down Backpropagation in Deep Feedforward Neural Networks
- 21 July 2025
- Published by Springer Nature
- p. 258-273
Top-Down Backpropagation in Deep Feedforward Neural Networks