Trustworthy and explainable AI

Making a model's behaviour inspectable and its use defensible: explainability, privacy-preserving learning, adversarial robustness, poisoning and backdoor attacks, unlearning, and design criteria for high-stakes decisions.

16 supervised theses. See all supervised theses for the full list, or open topics if you are looking for one to take on.

Level	Author	Title	Date	Industry	Recording
MSc	A. von Hertzen	Interpretability-Driven Regularization for Station-Level Weather Forecasting	May 2026		video
MSc	B. Prágai	Merkle Tree-based framework for verifiable, incentivized and privacy-preserving data contribution systems: A Proof-of-concept	Aug. 2026	Fixmeapp AB	video
MSc	I. Zalesskii	Privacy-Friendly Click-Through Rate Prediction in Programmatic Display Advertising	Aug. 2026	C Wire AG	video
MSc	B. Zheng	Exploring the structure in deep networks: Group, manifold and category theory	Dec. 2025
MSc	S. Rachidi	Explainable machine learning in cancer survival prediction — A case study of myeloma patients	July 2025
MSc	A. Ghazal	Zero-shot Machine Unlearning using GANs	July 2024	Nokia
BSc	A. Nguyen	Adversarial robustness of GPT-3.5 Turbo	Apr. 2024
MSc	M. Hilvo	The effect of privacy enhancing technologies on the quality of predictive models in healthcare - Cardiovascular disease models as an example	Oct. 2024	VTT
BSc	P. Frank	Explainable empirical risk minimization in high-stakes decision making	Sept. 2024
MSc	R. Mård	Backdoor attacks on large transformer-based regression model	Aug. 2024		video
MSc	R. Virtanen	Explainable AI Techniques in Trustworthy Object Detection	May 2024	TietoEVRY Oyj	video
BSc	S. Koivumaa	Design criteria for trustworthy AI	May 2024
BSc	I. Tulkki	Implementing backdoor data poisoning attacks	Apr. 2023
BSc	M. Lippo	Poisoning attacks on federated learning	Apr. 2023
BSc	R. Vikström	Trustworthy AI in drug development	Apr. 2023
BSc	S. Rachidi	Explainable machine learning in the prediction of breast cancer	July 2022