Thursday 03 May 2018: Gaussian processes: models and inference
Dr. Carl Henrik Ek - University of Bristol
Harrison 170 14:30-15:30
The science of machine learning is concerned with developing tools that provides means to integrate our prior assumptions with observed data. A natural formalism of this is Bayesian non-parametric which are methods that allow for unbounded model complexity and interpretable parametrisations. In this talk I will describe Gaussian and Dirichlet processes as examples of priors over infinite objects. I will describe how we can circmuvent the intractable inference by optimising a lower bound on the marginal likelihood. I will then exemplify the use of these model by showing recent work on with applications to airflow from windturbines and time-series data.
Dr. Carl Henrik Ek is a senior lecturer at the University of Bristol. His research focuses on developing computational models that allows machines to learn from data. In specific he is interested in Bayesian non-parametric models which allows for principled quantification of uncertainty, easy interpretability and adaptable complexity. He has worked extensively on models for representation learning with applications in robotics and computer vision.