Speaker
Petre Lameski
(Faculty of Computer Science and Engineering, University of Skopje, North Macedonia)
Description
The lecture covers the path from noisy data to scientifically defensible AI results, including signal-to-noise assessment, preprocessing, data leakage, validation, uncertainty, and reproducibility. It also presents explainable AI as a tool for auditing model behaviour while distinguishing model explanations from causal and scientific conclusions.