Xenia Miscouridou

Research

My research lies at the intersection of Bayesian statistics and machine learning, with interests in Bayesian nonparametrics, statistical network modelling, stochastic and point processes, and scalable Bayesian inference.

Bayesian modelling of networks

Visualization of a sparse network graph, with nodes coloured by community and sized by degree
Example network coloured by community, node size scaled by degree.

I develop Bayesian nonparametric models for sparse and heterogeneous networks, focusing on completely random measures, power-law degrees, overlapping communities, and networks that evolve over time. Current work includes dynamic, bipartite and covariate-dependent networks.

Bayesian nonparametrics · Completely random measures · Sparse random graphs · Dynamic networks · Community structure

Stochastic processes & spatio-temporal statistics

3D surface plot of a spatio-temporal intensity function over x and y coordinates
Posterior intensity surface from a spatio-temporal point-process model.

I develop flexible models for event data across space and time, particularly Hawkes and Cox processes, combining stochastic-process modelling with Gaussian-process and Bayesian nonparametric methods.

Hawkes processes · Cox processes · Gaussian processes · Point processes · Spatio-temporal modelling

Statistical machine learning & AI

Diagram of a feedforward neural network with an input layer, two hidden layers, and an output layer
Feedforward neural network architecture.

I work at the interface of machine learning and statistical methodology, including generative models, representation learning and interpretability. A newer direction explores neural and amortized methods for accelerating Bayesian inference while retaining principled uncertainty quantification.

Generative models · Interpretability · Neural processes · Amortized inference · AI-assisted Bayesian inference

Software

Open-source software and computational tools accompanying my research.