Guillaume Lajoie

This paper introduces hierarchical modular networks as reservoirs to study computational advantages of brain-like modularity, showing enhanced memory, multitasking, and temporal dynamics.

The paper shows that neural dynamics during movement are conserved across mice, monkeys, and humans, despite phylogenetic differences.

This study compares brain encoding performance of neural networks trained with different objectives on video game data, finding small advantages for reinforcement learning but overall brittle and limited generalization.

This study shows that leveraging the intrinsic manifold geometry of brain activity can accelerate learning of a noninvasive BCI by aligning brain activity within the manifold during task performance.

This paper uses artificial neural networks to study single-neuron adaptation mechanisms and finds they improve robustness and recover biological coding strategies like gain scaling.

This paper introduces exponentiated gradient (EG) as a learning algorithm for artificial neural networks that better aligns with biological principles than gradient descent.

This study shows that reinforcement learning produces neural dynamics in artificial networks that match biological motor cortex activity, suggesting RL underlies behavioral control development.

This paper proposes the oneirogen hypothesis, suggesting classical psychedelics induce hallucinations by shifting neural activity toward dream-like states, and validates this through simulations using a wake-sleep neural network model.

This paper uses neural biomarkers to improve the precision and efficacy of vagus nerve stimulation by analyzing inter-subject variability and optimizing dosing procedures.

The paper shows assistive decoders in BCIs influence neural representations by concentrating task-relevant information in fewer neurons and modes.

This paper develops a Bayesian optimization framework for real-time neural stimulation parameter optimization, achieving personalized dosing with fewer stimulations and uncertainty-aware safety.

conn2res is a Python toolbox that translates connectomes into artificial neural networks to study brain network computation.

This paper introduces T-PHATE, a nonlinear manifold learning algorithm for time-series data like fMRI, showing improvements in visualization, classification, and segmentation.

The paper investigates how different sites of plasticity affect adaptation to perturbations in brain-computer interfaces using a linear recurrent network.

This paper compares neural connectomes across 124 mammalian species using a unified protocol, finding that connectome organization aligns with traditional taxonomies.

This paper introduces a meta-learning framework to personalize neurostimulation parameters with fewer data points, demonstrating its effectiveness in vagus nerve stimulation.

This study shows how recurrent neural networks balance compression and expansion of activity dimensionality during learning to solve discrimination tasks effectively.

This paper uses human brain connectomes to train neuromorphic networks, showing that critical dynamics and modular organization enhance cognitive task performance.

This paper introduces a decision-making strategy called Performance-Gated Deliberation (PGD) that adapts to changing contexts by incorporating an adaptive opportunity cost of deliberation, outperforming traditional reinforcement learning models in explaining primate behavior and urgency.

This study investigates how predictive learning in neural networks generates low-dimensional latent representations of sensory inputs by analyzing network dynamics and intrinsic dimensionality.

This study investigates how anodal and cathodal tDCS affects cortical neuron activity in awake macaques, finding dose- and polarity-dependent changes in spike rates and network dynamics.

This paper introduces a hierarchical and adaptive optimization algorithm to efficiently map spatiotemporal neurostimulation patterns to motor outputs in non-human primates, revealing nonlinear computations in M1.