Joanna Marks

Hello ! I am a second-year PhD candidate in Statistics and Machine Learning at Imperial College London and the University of Oxford, where I am fortunate to be supervised by Deniz Akyildiz and Riccardo Passeggeri.

Before starting my PhD, I completed a BSc in Mathematics at the University of Warwick and then pursued an MSc in Statistics at Imperial College London which was possible thanks to a Bona Fide scholarship.

Research interests. My research focuses on problems at the intersection of generative modelling, particle methods, stochastic differential equations (SDEs), optimal transport and Schrodinger bridges. Recently I’ve been working on a project related to Gromov-Wasserstein distance computation. Always happy to chat!

news

Jun 15, 2026	I am excited to share that our joint work with Tim Y.J Wang and Deniz Akyildiz has been selected for an Oral presentation at UAI’26! I’m really grateful for the opportunity (only 2% of submitted papers were selected for oral presentations) and looking forward to presenting our work in Amsterdam!
Apr 02, 2026	Gabriel Rioux will present our joint work on “Gromov-Wasserstein distances between finite spaces: Duality, computation, and entropic approximation” with Riccardo Passeggeri at CDM-Seminars series at EPFL.Come chat to him if interested, pre-print coming soon!
Dec 02, 2025	I will be presenting a poster at Amortized ProbML workshop at ELLIS Unconference which will take place alongside EurIPS in Copenhagen. The poster will be a summary of our joint work with Tim Y.J Wang and Deniz Akyildiz on leveraging particle system for efficient training of Energy-Based Models. Come along if interested!

selected publications

UAI

Learning latent energy-based models via interacting particle Langevin dynamics

Joanna Marks, Tim YJ Wang, and O Deniz Akyildiz

2025

Oral Presentation Abs PDF Code

This paper was selected for an oral presentation at UAI 2026. Less than 2% of the submitted papers were chosen as oral presentations.

We develop interacting particle algorithms for learning latent variable models with energy-based priors. To do so, we leverage recent developments in particle-based methods for solving maximum marginal likelihood estimation (MMLE) problems. Specifically, we provide a continuous-time framework for learning latent energy-based models, by defining stochastic differential equations (SDEs) that provably solve the MMLE problem. We obtain a practical algorithm as a discretisation of these SDEs and provide theoretical guarantees for the convergence of the proposed algorithm. Finally, we demonstrate the empirical effectiveness of our method on synthetic and image datasets.