Dr. Prashant Singh (Uppsala University) -- Scalable learning under noise and uncertainty

11 December, 2025 15:00 (local Swedish time) Scalable Learning under Noise and Uncertainty Prashant Singh (Uppsala University) Abstract: Real-world data, particularly within scientific disciplines such as life sciences is often noisy, posing significant challenges for reliable machine learning. This talk formally introduces this ‘robust learning’ problem setting, presenting a Bayesian framework for training machine learning models that adapts automatically to varying noise levels in data. We will also approach another facet of noise, arising as prediction uncertainty from pre-trained models. We will characterize the nature of uncertainty structures within data and models, and motivate a principled uncertainty quantification pipeline.