Tutorial 3 2018
Machine Learning/Deep Learning in Remote Sensing
Abstract: Despite the wide and often successful application of machine learning techniques to analyse and interpret remotely sensed data, the complexity, special requirements, as well as selective applicability of these methods often hinders to use them to their full potential. The gap between sensor- and application-specific expertise on the one hand, and a deep insight and understanding of existing machine learning methods often leads to suboptimal results, unnecessary or even harmful optimizations, and biased evaluations. The aim of this tutorial is threefold: First, spread good practices for data preparation: Inform about common mistakes and how to avoid them (e.g. dataset bias, non-iid samples), provide recommendations about proper preprocessing and initialization (e.g. data normalization), and state available sources of data and benchmarks. Second, present efficient and advanced machine learning tools: Give an overview of standard machine learning techniques and when to use them (e.g. standard regression and classification techniques, clustering, etc.), as well as introducing the most modern methods, such as random fields, ensemble learning. Third, a particular focus will be put on deep learning. Central to the paradigm shift toward data-intensive science, deep learning has proven to be both a major breakthrough and an extremely powerful concept in many fields. The goal is to highlight what makes deep learning special in remote sensing, to showcase successful examples, to provide resources to make deep learning in remote sensing readily applicable, and more importantly, to discuss open issues.