\( \newcommand{\Ds}{\displaystyle} \newcommand{\PP}{{\mathbb P}} \newcommand{\RR}{{\mathbb R}} \newcommand{\KK}{{\mathbb K}} \newcommand{\CC}{{\mathbb C}} \newcommand{\ZZ}{{\mathbb Z}} \newcommand{\NN}{{\mathbb N}} \newcommand{\TT}{{\mathbb T}} \newcommand{\QQ}{{\mathbb Q}} \newcommand{\Abs}[1]{{\left|{#1}\right|}} \newcommand{\Floor}[1]{{\left\lfloor{#1}\right\rfloor}} \newcommand{\Ceil}[1]{{\left\lceil{#1}\right\rceil}} \newcommand{\sgn}{{\rm sgn\,}} \newcommand{\Set}[1]{{\left\{{#1}\right\}}} \newcommand{\Norm}[1]{{\left\|{#1}\right\|}} \newcommand{\Prob}[1]{{{{\mathbb P}}\left[{#1}\right]}} \newcommand{\Mean}[1]{{{{\mathbb E}}\left[{#1}\right]}} \newcommand{\cis}{{\rm cis}\,} \renewcommand{\Re}{{\rm Re\,}} \renewcommand{\Im}{{\rm Im\,}} \renewcommand{\arg}{{\rm arg\,}} \renewcommand{\Arg}{{\rm Arg\,}} \newcommand{\ft}[1]{\widehat{#1}} \newcommand{\FT}[1]{\left(#1\right)^\wedge} \newcommand{\Lone}[1]{{\left\|{#1}\right\|_{1}}} \newcommand{\Linf}[1]{{\left\|{#1}\right\|_\infty}} \)

Local times in Greece. All talks are 20 min + 10 min for questions.

The zoom link is https://zoom.us/j/99501476059. Please do not publish.

Speaker	Time
Pelekis	16:00 - 16:20
Karagiannis	16:30 - 16:50
Triantafyllou	17:00 - 17:20
Stamatakis	17:30 - 17:50
Alexopoulos	18:00 - 18:20

1. Angelos Alexopoulos: Bayesian variable selection for Gaussian copula regression models

   We develop a novel Bayesian method to select important predictors in regression models with multiple response of diverse types. In particular, a sparse Gaussian copula regression model is used to account for the multivariate dependencies between any combination of discrete and continuous responses and their association with a set of predictors. We utilise the parameter expansion for data augmentation strategy to construct a Markov chain Monte Carlo algorithm for the estimation of the parameters and the latent variables of the model. Based on a centred parametrization of the Gaussian latent variables, we design an efficient proposal distribution to update jointly the latent binary vectors of important predictors and the corresponding non-zero regression coefficients. The proposed strategy is tested on simulated data and applied to two real data sets in which the responses consist of low-intensity counts, binary, ordinal and continuous variables.

2. Georgios Karagiannis: Bayesian analysis of multifidelity computer models with local features and non-nested experimental designs

   We present a new method, the Augmented Bayesian Treed Co-Krigin, that extends the scope of co-kriging used for the statistical analysis of expensive computer models running at different fidelity levels. The proposed method, unlike existing ones, can take into account local features and discrepancies, as well as it can be used with non-nested experimental designs. Our procedure utilizes binary treed partition ideas that allow the representation of local features, and discovery of sudden changes in the multifidelity setting. To facilitate the parameter and predictive inference, we design a reversible jump MCMC sampler tailored to the proposed model. The good performance of our method is demonstrated on artificial benchmark examples, and compared against existing methods. The proposed method is implemented for the analysis of a large-scale climate modelling application with the Weather Research and Forecasting climate model.

3. Christos Pelekis: How to gamble, if you can afford it

   Suppose that you enter a casino with an amount of $\delta$ dollars, and with the hope of exiting from it with $t$ dollars, where $t > \delta$. There are $n$ mutually independent games you can gamble on, and the probability of winning each game equals $p \in (0,1)$. You divide your money into $n$ stakes, $\delta_1,...,\delta_n$, and lay a wager of $\delta_i$ dollars on the $i$-th game. The return from each game is $r$-fold, where $r > 1$. How should you choose the value of the stakes so that you maximise the probability of receiving at least $t$ dollars? I will introduce the problem and illustrate its relation to existing work on extremal set theory.

4. Marios Stamatakis: Hydrodynamic Limits of phase separating particle systems

   The objective of the theory of hydrodynamic limits is the macroscopic description of the thermodynamic characteristics of systems with a large number of interacting particles via an evolutionary partial differential equation, the so-called hydrodynamic equation. Since the end of the 20th century rigorous mathematical methods have been devised for obtaining the hydrodynamic equation of particle systems. However none of these methods is so far applicable to particle systems exhibiting phase transition. The main goal of our research is obtaining the hydrodynamic limit of phase separating particle systems. We focus on condensing Zero Range processes, which constitute a prototype model of particle systems exhibiting phase transition via the emergence of a condensate. Difficulties in obtaining their hydrodynamic limit arise both in the passage from the microscopic-stochastic level to the macroscopic-deterministic one and in the rigorous study of the expected hydrodynamic equation, which is a strongly degenerate quasilinear parabolic equation with two-phases. In this talk we will briefly present our progress towards obtaining the hydrodynamic equation of condensing Zero Range processes.

5. Sofia Triantafyllou: Causal discovery and inference from heterogeneous data sets

   Much of intelligent behavior involves causal inference, i.e., predicting the causal effects of actions. Probabilistic causal models are graphical models that connect causal properties of a system to probabilistic properties under observation and intervention. Under some conditions, these models allow the estimation of causal effects, and can be learned from data using statistical methods. My research focuses on learning causal models from multiple data sets, sampled from heterogeneous populations and conditions. In this talk, I will present of my recent work on learning causal models and estimating causal parameters from mixtures of experimental and observational data. I will show that integrating multiple datasets improves causal model learning and estimation of causal effects on simulated and real data.