Machine Learning for Planetary Science

"Machine Learning for Planetary Science presents planetary scientists with a way to introduce machine learning into the research workflow as increasingly large nonlinear datasets are acquired from planetary exploration missions. The book explores research that leverages machine learning methods to enhance our scientific understanding of planetary data and serves as a guide for selecting the right methods and tools for solving a variety of everyday problems in planetary science using machine learning. Illustrating ways to employ machine learning in practice with case studies, the book is clearly organized into four parts to provide thorough context and easy navigation. The book covers a range of issues, from data analysis on the ground to data analysis onboard a spacecraft, and from prioritization of novel or interesting observations to enhanced missions planning. This book is therefore a key resource for planetary scientists working in data analysis, missions planning, and scientific observation."-- Publisher website Front Cover Machine Learning for Planetary Science Copyright Contents Contributors Foreword References 1 Introduction to machine learning 1.1 Overview of machine learning methods 1.2 Supervised learning 1.2.1 Classification 1.2.2 Regression 1.3 Unsupervised learning 1.3.1 Clustering 1.3.2 Dimensionality reduction 1.4 Semisupervised learning 1.4.1 Self-training 1.4.2 Self-training with Expectation Maximization 1.4.3 Cotraining 1.5 Active learning 1.5.1 Uncertainty sampling 1.5.2 Query-by-committee 1.6 Popular machine learning methods 1.6.1 Principal component analysis 1.6.2 K-means clustering 1.6.3 Support vector machines 1.6.4 Decision trees and random forests 1.6.5 Neural networks 1.7 Data set preparation References 2 The new and unique challenges of planetary missions 2.1 Introduction 2.1.1 50 years of Mercury exploration 2.1.2 Challenges of large and complex data return 2.1.3 Facing the unknown 2.1.4 Machine learning for planetary science References 3 Finding and reading planetary data 3.1 Data acquisition in planetary science 3.1.1 Introduction 3.1.2 Data processing levels 3.1.3 PDS 3.1.3.1 Organizational structure within a node Releases and volumes EDR and RDR PDS4 collections and bundles 3.1.4 ESA's Planetary Science Archive 3.1.5 Reading data with Python 3.1.5.1 Example reading of PDS3 data 3.1.5.2 Troubleshooting data reading 3.1.6 Spaces to watch 3.1.6.1 PDR 3.1.6.2 PlanetaryPy 3.1.6.3 OpenPlanetary 4 Introduction to the Python Hyperspectral Analysis Tool (PyHAT) 4.1 Introduction 4.2 PyHAT library architecture 4.3 PyHAT orbital 4.3.1 Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) 4.3.2 Moon Mineralogy Mapper (M3) 4.3.3 Kaguya Spectral Profiler 4.4 PyHAT in-situ 4.4.1 Baseline removal example 4.4.2 Regression analysis example 4.4.3 Data exploration example 4.4.4 Calibration transfer 4.5 Conclusion Acronyms Acknowledgments References 5 Tutorial: how to access, process, and label PDS image data for machine learning 5.1 Introduction 5.2 Access to PDS data products 5.2.1 PDS Image Atlas 5.2.2 PDS Imaging Node Data Portal 5.3 Preprocessing PDS data products into standard image formats 5.3.1 PDS image data products 5.3.2 PDS browse images 5.3.3 Converting PDS image data products 5.4 Labeling image data 5.4.1 Publicly available labeled image data sets 5.4.2 Tools for labeling image data 5.5 Example PDS image classifier results 5.5.1 Train, validation, and test sets 5.5.2 Model fine-tuning 5.5.3 Model calibration and performance 5.5.4 Access to HiRISENet classification results 5.6 Summary Acknowledgments References 6 Planetary image inpainting by learning mode-specific regression models 6.1 Introduction 6.2 Related works 6.3 Experimental data 6.4 Proposed method 6.4.1 Unsupervised separation of histogram clusters 6.5 Network architecture 6.5.1 Training details 6.5.2 Reflection based information enhancement methodology 6.6 Experimental results 6.6.1 Performance evaluation 6.7 Conclusion References 7 Automated surface mapping via unsupervised learning and classification of Mercury Visible–Near-Infrared reflectance spectra 7.1 Introduction 7.2 Mercury and the MASCS instrument 7.3 Data preparation 7.4 Learning from multivariate data 7.4.1 Dimensionality reduction: ICA 7.4.2 Manifold learning 7.4.3 Cluster analysis 7.4.4 Conclusion Acknowledgment References 8 Mapping storms on Saturn 8.1 Introduction 8.1.1 Cassini-Huygens and ammonia clouds 8.2 Exploratory principal component analysis 8.3 A deep learning approach 8.3.1 Preprocessing and prelabeling 8.3.2 Neural network 8.3.3 Training and hyperparameter optimization 8.3.4 Classification validation 8.4 Saturn's feature map References 9 Machine learning for planetary rovers 9.1 Introduction 9.2 Risk- and resource-aware AutoNav 9.2.1 Overview 9.2.2 Terrain classification 9.2.3 Rock hazard detection 9.2.4 Vision-based slip and driving energy prediction 9.3 Drive-by-science 9.3.1 Overview 9.3.2 SCOTI: scientific captioning of terrain images 9.3.3 Image similarity search 9.3.4 DBS interface 9.3.5 DBS experiment with scientists 9.3.5.1 Methods 9.3.5.2 Tasks 9.3.5.3 Results 9.4 Demonstration on a test rover 9.5 Conclusion and future work References 10 Combining machine-learned regression models with Bayesian inference to interpret remote sensing data 10.1 The need for accurate fast-forward functions 10.2 Bayesian approach to inverse problems 10.3 Machine-learning-based surrogate models 10.4 Case study: constraining the thermal properties of asteroids with surrogate models 10.4.1 Dataset of thermophysical simulations 10.4.2 Surrogate infrared model of a mixture of regolith and rock 10.4.3 Bayesian inference of Itokawa's thermophysical properties 10.5 Future perspectives for data fusion 10.5.1 Remote sensing data fusion 10.5.2 Planet formation theory 10.5.3 Spacecraft autonomy References Index Back Cover