Statistical theory
Model selection, oracle inequalities, penalisation, minimal penalties, missing-data methodology, and non-asymptotic analysis for heterogeneous and high-dimensional models.
Research & Publications
Research & publications
My research program and its full publication record in one place: a research overview, then the grouped catalogue across 8 CV subsections covering 40 outputs. Many results are also available as interactive demonstrations.
Scalability and robustness in scientific statistical machine learningModel selectionApproximation capabilities and convergence rates of the mixture of experts modelsDeep neural networksSimulation-based inferenceAsymptotic statisticsBayesian nonparametricsMissing data
Filter
Catalogue
40 outputs grouped across 8 CV subsections.
Research overview
My research program sits at the intersection of mathematical statistics, machine learning, and scientific applications. I design methods that are theoretically justified, computationally scalable, and practically valuable. The full publication catalogue follows below; many of these results are available as interactive demonstrations.
Mixture-of-experts models
Approximation theory, parameter estimation, identifiability, scalable optimisation, sparse modelling, and uncertainty-aware inference for mixture-of-experts and related conditional mixture models.
Bayesian computation
Bayesian synthetic likelihood, surrogate posteriors, MCMC, variational methods, uncertainty quantification, and principled probabilistic inference under computational and modelling constraints.
AI for science
Applications in systems biology, genomics, transcriptomics, proteomics, and scientific machine learning, with a focus on interpretable and trustworthy methods for real scientific discovery.
Scalability and robustness in scientific statistical machine learning
Revisiting Incremental Stochastic Majorization-Minimization Algorithms with Applications to Mixture of Experts
2026 · arXiv preprint arXiv:2601.19811
Fast Model Selection and Stable Optimization for Softmax-Gated Multinomial-Logistic Mixture of Experts Models
2026 · arXiv preprint arXiv:2602.07997
Bayesian Parameter Balancing Enables Robust and Consistent Estimation of Kinetic Parameter Uncertainty
2026 · bioRxiv
StructSAM: Structure- and Spectrum-Preserving Token Merging for Segment Anything Models
2026 · ICML 2026 Workshop on AdaptFM: Resource-Adaptive Foundation Model Inference
Lambda-PSD: Scalable Approximate SNR-Optimised Polynomial Stein Discrepancies
2026 · ICML 2026 Workshop on Hypothesis Testing
StructSAM: Structure- and Spectrum-Preserving Token Merging for Segment Anything Models
2026 · arXiv preprint arXiv:2603.07307
Dendrograms of Mixing Measures for Softmax-Gated Gaussian Mixture of Experts: Consistency Without Model Sweeps
2026 · Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, AISTATS 2026 Spotlight Acceptance rate 2.5% over 2102 submissions.
Model Selection for Gaussian-gated Gaussian Mixture of Experts Using Dendrograms of Mixing Measures
2025 · arXiv preprint arXiv:2505.13052
Accelerating Transformers with Spectrum-Preserving Token Merging
2024 · Advances in Neural Information Processing Systems, NeurIPS 2024 Acceptance rate 25.8% over 15671 submissions.
HyperRouter: Towards Efficient Training and Inference of Sparse Mixture of Experts
2023 · Proceedings of the 2023 Empirical Methods in Natural Language Processing, EMNLP 2023 Main Acceptance rate 14% over 1041 submissions.
Model selection
Modifications of the BIC for order selection in finite mixture models
2026 · Annals of the Institute of Statistical Mathematics
On the large-sample limits of some Bayesian model evaluation statistics
2026 · Econometrics and Statistics
A non-asymptotic risk bound for model selection in high-dimensional mixture of experts via joint rank and variable selection
2023 · Australasian Joint Conference on Artificial Intelligence 2023, AJCAI 2023 Long Oral Presentation Acceptance rate 11% over 213 submissions.
A non-asymptotic approach for model selection via penalization in high-dimensional mixture of experts models
2022 · Electronic Journal of Statistics
Model selection by penalization in mixture of experts models with a non-asymptotic approach
2022 · JDS 2022 - 53èmes Journées de Statistique de la Société Française de Statistique (SFdS)
Model Selection and Approximation in High-dimensional Mixtures of Experts Models: From Theory to Practice
2021 · PhD Thesis, Normandie Université
An l1-oracle inequality for the Lasso in high-dimensional mixtures of experts models
2020 · arXiv preprint arXiv:2009.10622
Approximation capabilities and convergence rates of the mixture of experts models
Approximation rates for finite mixtures of location-scale models
2025 · arXiv preprint arXiv:2508.10612
Towards Convergence Rates for Parameter Estimation in Gaussian-gated Mixture of Experts
2024 · Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, AISTATS 2024 Acceptance rate 27.6% over 1980 submissions.
A General Theory for Softmax Gating Multinomial Logistic Mixture of Experts
2024 · Proceedings of the 41st International Conference on Machine Learning, ICML 2024 Acceptance rate 27.5% over 9473 submissions.
Risk Bounds for Mixture Density Estimation on Compact Domains via the h-Lifted Kullback-Leibler Divergence
2024 · Transactions on Machine Learning Research
Demystifying Softmax Gating Function in Gaussian Mixture of Experts
2023 · Advances in Neural Information Processing Systems, NeurIPS 2023 Spotlight Acceptance rate 3.6% over 12343 submissions.
Approximation of probability density functions via location-scale finite mixtures in Lebesgue spaces
2022 · Communications in Statistics - Theory and Methods
Approximations of conditional probability density functions in Lebesgue spaces via mixture of experts models
2021 · Journal of Statistical Distributions and Applications
Approximation by finite mixtures of continuous density functions that vanish at infinity
2020 · Cogent Mathematics & Statistics
Deep neural networks
FACET: A Fragment-Aware Conformer Ensemble Transformer
2026 · The Fourteenth International Conference on Learning Representations, ICLR 2026 Acceptance rate 28% over close to 19000 submissions.
From Fragments to Geometry: A Unified Graph Transformer for Molecular Representation from Conformer Ensembles
2025 · ICML 2025 Generative AI and Biology (GenBio) Workshop
ExGra-Med: Extended Context Graph Alignment for Medical Vision-Language Models
2025 · NeurIPS 2025 (Advances in Neural Information Processing Systems 38)
Enriched Instruction-Following Graph Alignment for Efficient Medical Vision-Language Models
2025 · ICML 2025 Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences
CompeteSMoE--Effective Training of Sparse Mixture of Experts via Competition
2024 · arXiv preprint arXiv:2402.02526
Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks
2024 · Proceedings of the 41st International Conference on Machine Learning, ICML 2024 Acceptance rate 27.5% over 9473 submissions.
Simulation-based inference
Revisiting Concentration Results for Approximate Bayesian Computation
2025 · Bayesian Analysis
Bayesian Likelihood Free Inference using Mixtures of Experts
2024 · International Joint Conference on Neural Networks, IJCNN 2024 Acceptance rate 52% over 3272 submissions.
Mixture of expert posterior surrogates for approximate Bayesian computation
2022 · JDS 2022 - 53èmes Journées de Statistique de la Société Française de Statistique (SFdS)
Summary statistics and discrepancy measures for approximate Bayesian computation via surrogate posteriors
2022 · Statistics and Computing
Approximate Bayesian computation with surrogate posteriors
2021 · ISBA 2021 - World Meeting of the International Society for Bayesian Analysis
Asymptotic statistics
On the Asymptotic Distribution of the Minimum Empirical Risk
2024 · Proceedings of the 41st International Conference on Machine Learning, ICML 2024 Acceptance rate 27.5% over 9473 submissions.
Bayesian nonparametrics
Bayesian nonparametric mixture of experts for inverse problems
2024 · Journal of Nonparametric Statistics
Bayesian nonparametric mixture of experts for high-dimensional inverse problems
2022 · BNP13 – 13th Conference on Bayesian Nonparametrics
Missing data
A Unified Framework for Variable Selection in Model-Based Clustering with Missing Not at Random
2025 · Advances in Neural Information Processing Systems, NeurIPS 2025 Acceptance rate 24.52% over 21575 submissions.