A UW course for Machine Learning for the Geoscience
Instructor: Marine Denolle (mdenolle at uw.edu) and Akshay Mehra (akmehra at uw.edu)
Supported by the GeoSMART team (Stefan Todoran, Nicoleta Cristea, Anthony Arendt, Scott Henderson, Ziheng Sun)
The course is intended to introduce Machine Learning in Geosciences, the basics of computing, and methodologies in applied machine learning. The course focuses on canonical and topical data sets in seismology, oceanography, cryosphere, planetary sciences, geology, and geodesy. The methods taught include unsupervised clustering, logistic regression, random forest, support vector machine, and deep learning.
By the end of the quarter, the students should be able to:
- Demonstrate computing skills in python, jupyter notebooks, Git version control, and deploy scripts on local computers, cloud-hosted hubs, or cloud instances.
- Develop and apply standard machine-learning workflows: 1) Data preparation, 2) Model design, 3) Model training, validation, and evaluation.
- Apply standard data manipulation strategies in the Geosciences: data types (time series and geospatial), data formats, data visualization, dimensionality reduction, and feature engineering.
- Describe and demonstrate the adoption of open science principles, science reproducibility, and digital scholarship.
- Describe the canonical examples in a breadth of disciplines in geoscience.
- Understand at least qualitatively how some of the advanced techniques (Fourier and wavelet transform, principal component analysis, โฆ) manipulate and transform the data to interpret the output.
This course is being developed in conjunction with the GeoSMART curriculum book
Prerequisites: MATH 207 and MATH 208, or MATH 307 or 308, or AMATH 351 or 352, CS160 or CS163, or permission from the instructor.
Recommended: Knowledge in Matlab or python, AMATH301, 100- or 200-level courses in the Earth Sciences. Refreshers in computing skills will be provided.
- Module 1 (weeks 1 and 2) Open-GeoScience Ecosystem
- Module 2 (weeks 3 and 4) ML Ready Data Set
- Module 3 (weeks 5-6-7) Machine Learning
- Module 4 (weeks 8-9-10) Deep Learning
Each week, students will write a short report about either a paper or a webinar. Use the template on canvas and answer the questions when appropriate. Submissions of the report PDF are due Wednesdays at 11:59 pm PDT on canvas. The instructor will spend 15 minutes Monday morning summarizing the reading and webinar reports. Papers can be found and/or uploaded on a shared private course Google Drive here
Good luck!