Gaussian Processes are a statistical popularized by the Machine Learning community. They provide several advantages over more widespread techniques like Neural Networks (Deep Learning), Support Vector Machines or Ensemble Learning. Gaussian Processes allow to draw statistical levels of confidence over their inference (prediction) and, unlike mainstream Machine Learning techniques, can make accurate predictions with small datasets. This enables Gaussian Processes to be useful in the context of many science and engineering applications.
This minicourse will be divided into two sections: 1) the first section will introduce Gaussian Processes, their construction and derivation, paying particular attention to the intuition behind them. 2) the second section will focus on using them for various applications, including Bayesian optimization and modelling of stochastic dynamic systems.
Python codes and notes will be provided so that the attendants can explore the codes and implementations on their own after the tutorial.
Author: Prof. Antonio del Rio Chanona is a Research Fellow at the Centre for Process Systems Engineering, Imperial College London, developing and applying computer algorithms from the area of optimization and machine learning to engineering systems.