Dr. Kellian Cottart

Publications

Recent published work, with papers, materials and citations.

5 papers: 4 in journals, 1 at conferences. Also on Google Scholar and ORCID.

HighlightICML 2026, my latest and favourite contribution

Active Continual Learning with Metaplastic Binary Bayesian Neural Networks

We introduce BiMU, a Bayesian update rule for binary neural networks that prevents posterior saturation on long non-stationary streams, sustaining epistemic uncertainty and enabling buffer-free active querying to cut label and backpropagation costs.

K. Cottart, T. Ballet, D. Bonnet, D. Querlioz

BibTeX 

@inproceedings{cottart2026active,
  title     = {Active Continual Learning with Metaplastic Binary
               Bayesian Neural Networks},
  author    = {Cottart, Kellian and Ballet, Th\'eo and
               Bonnet, Djohan and Querlioz, Damien},
  booktitle = {Proceedings of the 43rd International Conference
               on Machine Learning},
  series    = {PMLR},
  volume    = {306},
  year      = {2026},
  eprint    = {2605.30198},
  archivePrefix = {arXiv}
}

All publications

all as .bib

2026

  1. Appl. Phys. Lett.

    Differential Magneto-Ionic Artificial Synapse

    Two magneto-ionic devices are driven differentially, so the synaptic weight is their conductance difference: bipolar weights from unipolar devices, and drift common to both devices cancels out.

    H. N. Mohanty, G. Bernard, M.-A. Syskaki, K. Cottart, A. Durnez, J. Langer, D. Querlioz, L. Herrera Diez

    Journal article, volume 128, issue 23, article 232404

    BibTeX 

    @article{mohanty2026differential,
  title   = {Differential magneto-ionic artificial synapse},
  author  = {Mohanty, H. N. and Bernard, G. and Syskaki, M.-A. and Cottart, K. and
             Durnez, A. and Langer, J. and Querlioz, D. and Herrera Diez, L.},
  journal = {Applied Physics Letters},
  volume  = {128},
  number  = {23},
  pages   = {232404},
  year    = {2026},
  doi     = {10.1063/5.0334517}
}

  2. ICML

    Active Continual Learning with Metaplastic Binary Bayesian Neural Networks · also above

    We introduce BiMU, a Bayesian update rule for binary neural networks that prevents posterior saturation on long non-stationary streams, sustaining epistemic uncertainty and enabling buffer-free active querying to cut label and backpropagation costs.

    K. Cottart, T. Ballet, D. Bonnet, D. Querlioz

    Conference paper, Proceedings of Machine Learning Research 306

    BibTeX 

    @inproceedings{cottart2026active,
  title     = {Active Continual Learning with Metaplastic Binary
               Bayesian Neural Networks},
  author    = {Cottart, Kellian and Ballet, Th\'eo and
               Bonnet, Djohan and Querlioz, Damien},
  booktitle = {Proceedings of the 43rd International Conference
               on Machine Learning},
  series    = {PMLR},
  volume    = {306},
  year      = {2026},
  eprint    = {2605.30198},
  archivePrefix = {arXiv}
}

2025

  1. Nature Communications

    Bayesian Continual Learning and Forgetting in Neural Networks

    We introduce Metaplasticity from Synaptic Uncertainty (MESU), a Bayesian update rule that scales each parameter’s learning by its uncertainty, enabling a principled combination of learning and forgetting without explicit task boundaries.

    D. Bonnet, K. Cottart, T. Hirtzlin, T. Januel, T. Dalgaty, E. Vianello, D. Querlioz

    Journal article, volume 16, article 9614

    BibTeX 

    @article{bonnet2025bayesian,
  title   = {Bayesian continual learning and forgetting
             in neural networks},
  author  = {Bonnet, Djohan and Cottart, Kellian and Hirtzlin, Tifenn and
             Januel, Tarcisius and Dalgaty, Thomas and Vianello, Elisa and
             Querlioz, Damien},
  journal = {Nature Communications},
  volume  = {16},  number = {1},  pages = {9614},
  year    = {2025},  doi = {10.1038/s41467-025-64601-w}
}

  2. Nano Letters

    Dynamic Control of Weight-Update Linearity in Magneto-Ionic Synapses

    We demonstrate that magneto-ionic devices can perform as synaptic elements with dynamically tunable depression linearity controlled by an external magnetic field, a functionality reminiscent of neuromodulation in biological systems.

    G. Bernard, K. Cottart, M.-A. Syskaki, V. Porée, A. Resta, A. Nicolaou, A. Durnez, S. Ono, A. Mora Hernandez, J. Langer, D. Querlioz, L. Herrera Diez

    Journal article, volume 25, issue 4, 1443–1450

    BibTeX 

    @article{bernard2025dynamic,
  title   = {Dynamic Control of Weight-Update Linearity
             in Magneto-Ionic Synapses},
  author  = {Bernard, Guillaume and Cottart, Kellian and Syskaki, Maria-Andromachi and
             Por{\'e}e, Victor and Resta, Andrea and Nicolaou, Alessandro and
             Durnez, Alan and Ono, Shimpei and Mora Hernandez, Ariam and
             Langer, Juergen and Querlioz, Damien and Herrera Diez, Liza},
  journal = {Nano Letters},
  volume  = {25},  number = {4},  pages = {1443--1450},
  year    = {2025},  doi = {10.1021/acs.nanolett.4c05247}
}

2024

  1. J. Composite Materials

    A Unified Method to Generate Representative Volume Elements with Tailored Random Fibre Arrangements to Estimate the Shear and Transverse Behaviours of Unidirectional Continuous Fibres Composite Plies

    Different microstructures with or without clusters of fibres (resulting from manufacturing processes) are necessary to quantify the mechanical variability. The variability of the microstructures is characterised by the variation of the arrangement, the volume fraction and the misalignment of the fibres.

    P.-Y. Mechin, A. Borras, K. Cottart, V. Keryvin

    Journal article, online first, doi 10.1177/00219983241300144

    BibTeX 

    @article{mechin2024unified,
  title   = {A Unified Method to Generate Representative Volume
             Elements with Tailored Random Fibre Arrangements to
             Estimate the Shear and Transverse Behaviours of
             Unidirectional Continuous Fibres Composite Plies},
  author  = {Mechin, P.-Y. and Borras, A. and Cottart, K. and
             Keryvin, V.},
  journal = {Journal of Composite Materials},
  year    = {2024},
  doi     = {10.1177/00219983241300144}
}

Older items, co-authored abstracts and posters are on Google Scholar and ORCID; the CV has the full record, including talks and teaching.