On Friday, November 14, 2025, at 12.15 pm, ms Elianna Douka, student of the postgraduate program “Data Science and Information Technologies”, will present his MSc thesis titled: “On evaluating counterfactual explanations for Machine Learning Models”

Titlle: “On evaluating counterfactual explanations for Machine Learning Models”

Elianna Douka, DSIT MSc

Abstract
The need to interpret machine learning (ML) model outputs using counterfactual (CF) explanations, small, meaningful perturbations of input data, has gained increasing importance in the research community. While CF explanations provide valuable insight into model behavior, evaluating their performance remains challenging, especially when models are trained in distributed or federated settings. This work draws inspiration from interactive system evaluation and introduces a set of metrics designed to assess CF explanations in distributed learning environments. The proposed approach aims to ensure that CF examples contribute to unbiased, stable, and interpretable model training across decentralized data sources.
Preliminary experimental results demonstrate the effectiveness and robustness of the proposed metrics, highlighting their potential to improve model fairness, reliability, and transparency in distributed machine learning scenarios.
Examiners
Prof. M. Koubarakis,
Prof. V. Kalogeraki

link
https://uoa.webex.com/uoa/j.php?MTID=m593ef98086eb56ade69da482acb33294