Research interests

My research interests lie in the field of computational Bayesian inference, with a focus on developing hardware-specialised methods, for example, designing sampling algorithms that explicitly take advantage of the parallelism of modern GPUs rather than implicitly. Other interests lie in embarrassing (or not) parallelisation of statistical algorithms, statistical efficiency in state-space models, gradient-based inference, coupling methodologies, and optimal transport.

Publications

Below is a list of my publications, and a list of my working papers.

Working papers

A. Cabezas, A. Corenflos, J. Lao, R. Louf et al. (2024): BlackJAX: Composable Bayesian inference in JAX. arXiv

S. Iqbal, A. Corenflos, S. Särkkä, H. Abdulsamad (2024): Nesting Particle Filters for Experimental Design in Dynamical Systems. under review, arXiv

A. Corenflos, A. Finke (2024): Particle-MALA and Particle-mGRAD: Gradient-based MCMC methods for high-dimensional state-space models. under review, arXiv

Y. Le Fay, S. Särkkä, A. Corenflos (2023): Modelling pathwise uncertainty of Stochastic Differential Equations samplers via Probabilistic Numerics. under review, arXiv

N. Bosch, A. Corenflos, F. Yaghoobi, F. Tronarp, P. Hennig, S. Särkkä (2023): Parallel-in-Time Probabilistic Numerical ODE Solvers. under review, arXiv

H. Abdulsamad, S. Iqbal, A. Corenflos, S. Särkkä, (2024): Risk-Sensitive Stochastic Optimal Control as Rao-Blackwellized Markovian Score Climbing. under review, arXiv

A. Corenflos, M. Sutton, N. Chopin (2023): Debiasing Piecewise Deterministic Markov Process samplers using couplings. work in progress, arXiv

A. Corenflos, S. Särkkä (2023): Auxiliary MCMC and particle Gibbs samplers for parallelisable inference in latent dynamical systems. under review, arXiv

F. Yaghoobi, A. Corenflos, S. Hassan, S. Särkkä (2022): Parallel square-root statistical linear regression for inference in nonlinear state space models. under review, arXiv

Published/Accepted papers

A. Corenflos, H. Abdulsamad (2023): Variational Gaussian filtering via Wasserstein gradient flows EUSIPCO 2023 arxiv

A. Corenflos, N. Chopin, S. Särkkä (2023): De-Sequentialized Monte Carlo: a parallel-in-time particle smoother JMLR 23(283). open access

A. Corenflos, Z. Zhao, S. Särkkä (2022): Temporal Gaussian Process Regression in Logarithmic Time. FUSION. official, arXiv

Rémi Flamary, Nicolas Courty, Alexandre Gramfort, et al.: “POT: Python Optimal Transport” JMLR 22(78). open access

F. Yaghoobi, A. Corenflos, S. Hassan, S. Särkkä (2021): Parallel Iterated Extended and Sigma-point Kalman Smoothers. ICASSP. official, arXiv

A. Corenflos, J. Thornton (joint first authorship), G. Degligiannidis, A. Doucet (2021): Differentiable Particle Filtering via Entropy-Regularized Optimal Transport ICML. open access

Talks

A. Corenflos (2023): Debiasing PDMP samplers. 41st Finnish Summer School on Probability and Statistics

A. Corenflos (2023): Auxiliary MCMC and particle Gibbs samplers for parallelisable inference in latent dynamical systems. Bayes comp’, session on state-space modelling and particle filtering link

A. Corenflos (2022): Auxiliary Kalman samplers for parallel inference in latent dynamical models. Workshop on Computational methods for unifying multiple statistical analyses link

A. Corenflos (2022): The Coupled Rejection Sampler. The Royal Statistical Society’s Young Statisticians Meeting 2022 - online, mathematical statistics session link

A. Corenflos (2022): The Coupled Rejection Sampler. Seminar on Advances in Probabilistic Machine Learning, Aalto Universtiy & ELLIS Helsinki, Finland Link

Posters

A. Corenflos (2021): Temporal Gaussian Process Regression in Logarithmic Time. Finnish AI Day link

Stashed papers

If you see something you like, have ideas, or would like to discuss, feel free to contact me. I’d be happy to reopen these projects if there is interest.

A. Corenflos, S. Särkkä (2022): The Coupled Rejection Sampler. arXiv