University of Bonn
Parallel and Scalable Machine Learning for Remote Sensing Big Data
WS 2019-2020
Abstract
This course exposes the student to the physical principles underlying satellite observations of Earth by passive sensors, as well as parallel and scalable machine (deep) learning algorithms for the automatic classification of land cover classes from remote sensing images.
Lecture 1 - Prologue
With a daily publication rate of over 26,500 products/day, and an average daily download volume of 166 TB, 🛰️ESA's Sentinel mission has been providing remote sensing applications with a number of challenges:
💻 Consumer-grade PCs lack the capacity to process the raw data
🖧 ML has high demands on memory and bandwidth throughput
⌚️ Time to solution is extensive
High Performance Computing modular architectures solve these issues by providing:
✔️ Sufficient memory and a high number of cores
✔️ Near-real-time processing (depending on the application)
This and more in my new course on Scalable Machine Learning for Remote Sensing Big Data. Here's the intro to the course 👇
Lecture 2 - Remote Sensing Systems
Outline
🛰️ The Remote Sensing Process
💡 Electromagnetic Spectrum and Radiation
− Wave and Quantum Model
− Planck’s and Stefan–Boltzmann Law
− Wien’s Displacement
⚡️🌍 Energy Interactions
− With the Atmosphere and Earth Surfaces
📸Platforms and Sensors
🚀 Earth Observation Missions
Lecture 3 - Machine Learning for Classification
Outline
🌄Land Use and Land Cover Classes
− USGS and CORINE
📈Spectral Reflectance
− Informative and Spectral Classes
− Spectral Response
🔎Patterns Recognition with Photointerpretation
🌁Pattern Recognition Systems
− Feature Extraction and Selection
− Automatic Classification
− Bayesian Decision
− Theory Artificial Neural Networks (ANNs)
− Convolutional Neural Networks (CNNs)
− 1D, 2D and 3D Convolutional Layers
− Model Evaluation
Lecture 4 - Parallel and Scalable Machine Learning with High Performance Computing
Outline
🍴The Free Lunch is Over
− Moore’s Law
− Work Harder and work smarter
− Get Help: Many-Core Era
⛓Hardware Levels of Parallelism
− In-core, In-Processor, Single and Multiple Computers
− Graphics Processing Units (GPUS)
📱High Performance Computing (HPC)
− TOP500
− Architectures of HPC Systems
⚙️Modular Supercomputing Architecture
− Deep projects
Previous teaching
