Generating mock matter density fields conditioned on primordial density fields with flow matching.
The diffusion process showing both the stochastic and deterministic paths through the marginal distributions of the diffusion process for a set of datapoints.
I like using generative models in Bayesian inference problems to extract information on fundamental physics in cosmology.
Right now i'm working on
- a simulation-based inference (SBI) package in
jax, - field-level inference pipelines using generative models.
I'm interested in generative models...
- Score-based diffusion models,
- Variational Diffusion models (VDMs),
- Variational Diffusion Autoencoders (VDAEs).
...transformer models & geometric deep learning...
...and statistical problems in general
- Frequentist-matching priors,
- Information-maximising neural networks,
- Hierarchical Bayesian Neural Networks (HBNNs).
I also teach MSc Physics students in the Physik x AI labs at LMU Physik where I write teaching material that delivers machine learning insights from problems in physics.
My goal is to show students cutting edge algorithms and statistical methods that they will not learn anywhere else.
- Simulation-based inference with random fields in cosmology,
- Generative models for Bayesian inference in cosmology (link not available).