Christos Nikolopoulos

Christos Nikolopoulos, Michael Eden, Adrian Muntean

We study a reaction-diffusion model posed on two distinct spatial scales that accounts for diffusion, aggregation, fragmentation, and deposition of populations of colloidal particles within a porous material. In this model, the macroscopic transport of the particles is described by an effective equation whose transport coefficients are determined by cell problems posed on the underlying pore scale. The internal pore geometry can change over time due to deposition or detachment of colloidal particles. We represent the evolving microstructure as solid cores whose phase boundaries can grow or shrink over time. As deposition progresses, neighbouring growing cores may come into contact, leading to local clogging of the pore space. We investigate how such evolving microstructures influence the effective transport and storage properties of porous layers. We establish basic analytical results concerning the weak solvability of the resulting multiscale evolution problem, which takes the form of a strongly non-linear parabolic system, in the non-clogging regime. For the numerical approximation of weak solutions we propose a two-scale finite element discretization. Numerical experiments illustrate how local clogging affects the effective dispersion tensor and quantify the resulting trade-off between transport efficiency and storage capacity.

Spyridon Evangelatos, Eleni Veroni, Vasilis Efthymiou et al.

Counterfactual explanations have emerged as a prominent method in Explainable Artificial Intelligence (XAI), providing intuitive and actionable insights into Machine Learning model decisions. In contrast to other traditional feature attribution methods that assess the importance of input variables, counterfactual explanations focus on identifying the minimal changes required to alter a model's prediction, offering a ``what-if'' analysis that is close to human reasoning. In the context of XAI, counterfactuals enhance transparency, trustworthiness and fairness, offering explanations that are not just interpretable but directly applicable in the decision-making processes. In this paper, we present a novel framework that integrates perturbation theory and statistical mechanics to generate minimal counterfactual explanations in explainable AI. We employ a local Taylor expansion of a Machine Learning model's predictive function and reformulate the counterfactual search as an energy minimization problem over a complex landscape. In sequence, we model the probability of candidate perturbations leveraging the Boltzmann distribution and use simulated annealing for iterative refinement. Our approach systematically identifies the smallest modifications required to change a model's prediction while maintaining plausibility. Experimental results on benchmark datasets for cybersecurity in Internet of Things environments, demonstrate that our method provides actionable, interpretable counterfactuals and offers deeper insights into model sensitivity and decision boundaries in high-dimensional spaces.