News

Queen Mary Statistics Symposium - Wed Dec 2nd

Centre for Probability, Statistics and Data Science 

20 November 2026

Wednesday 2 December, 10:00–16:00
Room MB-503, Mathematical Sciences Building
Queen Mary University of London
Mile End Road, London E1 4NS

The Queen Mary Statistics Symposium replaces our weekly Statistics and Data Science Seminar. Once each term we hold a one-day symposium where four speakers give talks, with food and conversation in between. We want more time to talk with the speakers and with each other, so please come for the whole day if you can, including lunch.

The event is in person only. Lunch and coffee are provided. Attendance is by registration through this form.

Programme

Time
10:00–10:30
Welcome coffee
10:30–11:20
Masha Naslidnyk (UCL)
11:20–12:10
Shifan Yu (University of Oxford)
12:10–13:30
Lunch
13:30–14:20
Chieh-Hsi Wu (University of Southampton)
14:20–15:10
Antonio Canale (University of Genova, currently visiting Glasgow)
15:10–16:00
Coffee and discussion

Titles and abstracts

Masha Naslidnyk (UCL)
Kernel Quantile Embeddings and Associated Probability Metrics
Embedding probability distributions into reproducing kernel Hilbert spaces (RKHS) has enabled powerful non-parametric methods such as the maximum mean discrepancy (MMD), a statistical distance with strong theoretical and computational properties. At its core, the MMD relies on kernel mean embeddings (KMEs) to represent distributions as mean functions in RKHS. However, it remains unclear if the mean function is the only meaningful RKHS representation. Inspired by generalised quantiles, we introduce the notion of kernel quantile embeddings (KQEs), along with a consistent estimator. We then use KQEs to construct a family of distances that: (i) are probability metrics under weaker kernel conditions than MMD; (ii) recover a kernelised form of the sliced Wasserstein distance; and (iii) can be efficiently estimated with near-linear cost. Through hypothesis testing, we show that these distances offer a competitive alternative to MMD and its fast approximations. Our findings demonstrate the value of representing distributions in Hilbert space beyond simple mean functions, paving the way for new avenues of research.

Shifan Yu (University of Oxford)
Realized Regularized Regressions
We develop a continuous-time penalized regression framework for the estimation of time-varying coefficients and variable selection when both the response and covariates are Itô semimartingales with jumps. The coefficient paths are approximated by spline basis expansions and estimated via least squares from truncated high-frequency increments. In a finite-dimensional setting, we establish consistency and derive a feasible asymptotic distribution for the integrated coefficient estimator under infill asymptotics. We then extend the framework to high-dimensional settings in which the number of candidate covariates diverges, and show that a group-wise penalized estimator with a truncated L1-penalty attains the oracle property, which delivers both consistent model selection and coefficient estimation. An empirical application to a large panel of more than two hundred high-frequency asset-pricing factors documents sparse factor structure across a large cross-section of stocks and industry portfolios.

Chieh-Hsi Wu (University of Southampton)
Bayesian Phylogenetics: Modelling, Evaluation and Application
A phylogenetic model assumes that the evolutionary relationships among individuals in a sample can be represented by a tree-like structure, a phylogenetic tree, which can be feasibly inferred from genetic data using Bayesian inference. By specifying appropriate priors on the tree, additional parameters can be estimated to provide insights into population dynamics and migration history.
This talk presents an integrated approach to estimating the population and migration dynamics in virus populations. Within this framework, common prior assumptions are made about the smoothness of trends in the population dynamics of interest. This talk also challenges these assumptions by presenting an information-theoretic technique that assesses the influence of smoothing priors, helping avoid overconfident and potentially misleading inferences.

Antonio Canale (University of Genova, currently visiting Glasgow)
Advances in Sparse Functional Data Analysis
Functional data often exhibit complex local structures that cannot be adequately captured by traditional global modeling assumptions. In this talk, I will discuss two recent and related methodological advances that exploit different forms of local sparsity to enhance flexibility and interpretability in functional data analysis. Both approaches leverage the local support property of B-spline basis functions, albeit in different ways.
The first contribution concerns function-on-function regression, where we introduce a locally sparse model allowing the regression coefficient surface to be exactly zero over selected regions of its domain. By exploiting structured penalization, the proposed approach identifies localized relationships between functional predictors and responses, bridging the gap between concurrent and fully nonconcurrent regression models.
The second contribution focuses on Bayesian local clustering of functional data, allowing curves to share common behaviors over portions of their domains without imposing a global clustering structure. This is achieved through B-spline representations and a novel class of semi-Markovian dependent random partitions, which capture the evolving structure of local similarities across functions.
The proposed methodologies are illustrated through simulation studies and applications to energy market dynamics and tide-level data.

People: Nicolás HERNáNDEZ

Updated by: Alexander Shestopaloff