SWEN Talks, 23 and 24 September: sustainable microservices and fairness in multi-class classification

The "SWEN Talks: Perspectives in Software Engineering" series continues next week with two seminars in the Alan Turing Seminar Room (DISIM). Both are open to students, researchers and practitioners; feel free to share the announcement with colleagues who may be interested.

Wednesday 23 September, 14:30-15:30

What We Know vs What We Do: A Research and Practitioner View on Sustainable Microservices by Eoan O'Dea, PhD student at SWEN, Università degli Studi dell'Aquila.

The ICT sector accounts for a significant and growing share of global electricity consumption, and microservices have become the default architectural style for cloud-native systems, adopted by the large majority of engineering organisations. Yet how sustainable these architectures actually are, in research and in practice, remains poorly understood. This talk reviews what the literature currently knows about energy consumption in microservice systems, drawn from a systematic literature review of 40 primary studies, and contrasts it with how sustainability is actually understood and practised in industry, based on a mixed-methods survey of 62 practitioners. The two pictures often diverge: technical knowledge and intent are rarely the limiting factor. Sustainability in industry turns out to be technically executed but organisationally gated, with adoption blocked less by missing tools or expertise than by the absence of organisational mandate, incentive, or measurement culture.

Thursday 24 September, 12:00-13:00

Constrained Optimisation for Fairness in Multi-Class Classification by Maryam Boubekraoui, Lecturer and Researcher at HESTIM (Morocco) and Postdoctoral Fellow at Cadi Ayyad University (Morocco), visiting researcher at SWEN.

The growing deployment of machine learning in sensitive domains such as healthcare, finance, education, and criminal justice has made algorithmic fairness a central concern: biased classifiers can systematically disadvantage individuals based on protected attributes like gender, race, or age. Correcting this without sacrificing predictive accuracy remains largely open, especially beyond simple binary decisions. Bias mitigation methods are generally grouped into three families depending on when they act in the ML pipeline: pre-processing approaches, which correct the training data before learning; post-processing approaches, which adjust an already-trained model's predictions; and in-processing approaches, which embed the fairness objective directly into the training process itself. Most existing methods, of all three kinds, were designed only for binary classification, leaving multi-class settings underexplored despite their frequent use in fields like education and healthcare. In this talk, I will present Generalised Exponentiated Gradient (GEG), a new method that, as an in-processing approach, has the advantage of directly shaping the model's decision boundary during learning rather than correcting it after the fact. GEG extends the reductions approach of Agarwal et al., originally formulated as a convex-concave saddle-point game between a learner and an auditor, to handle (i) multi-class label spaces and (ii) multiple fairness definitions enforced simultaneously, while preserving convexity and convergence guarantees. I will present extensive numerical results on ten real-world datasets, benchmarked against six baseline methods, showing that GEG consistently improves fairness with little to no loss in accuracy.

The method is described in M. Boubekraoui, G. D'Aloisio, A. Di Marco, "A Generalised Exponentiated Gradient Approach to Enhance Fairness in Binary and Multi-class Classification Tasks" (arXiv preprint, 2026), and builds on A. Agarwal et al., "A Reductions Approach to Fair Classification" (ICML 2018).

The full programme of past and upcoming seminars is on the SWEN Talks page.