140.650.01
AI Methods for Geospatial Data
Location
East Baltimore
Term
4th Term
Department
Biostatistics
Credit(s)
3
Academic Year
2024 - 2025
Instruction Method
In-person
Tu, Th, 3:30 - 4:50pm
Auditors Allowed
Yes, with instructor consent
Available to Undergraduate
No
Grading Restriction
Letter Grade or Pass/Fail
Course Instructor(s)
Contact Name
Frequency Schedule
Every Other Year
Resources
This course is timely and essential due to the increasing reliance on geospatial data analysis across various fields like climate science, epidemiology, and urban planning. There is a lack of a comprehensive course encompassing both statistical and machine learning paradigms tailored specifically for geospatial analysis. This course fills a critical gap in our curriculum, ensuring that learners acquire a comprehensive skill set tailored to the demands of contemporary geospatial data analysis.
Introduces advanced AI methods for analyzing large-scale geospatial data, with particular emphasis on geostatistics. Starts with an overview of geospatial data and exploratory data analysis and visualization techniques. Delves next into Gaussian Processes (GP) and kriging for spatial data modeling, covering both classical optimization and Bayesian MCMC methods for GP models. Covers spatial graphical models and Nearest Neighbor Gaussian Processes (NNGP) for handling massive datasets. Introduces machine learning techniques, including random forests and neural networks (multi-layer perceptrons, convolutional and graph neural networks), and explores hybrid methods that combine traditional statistical modeling with machine learning. Covers various state-of-the-art computational techniques, like stochastic approximations and variational Bayesian optimization, and offers a hands-on demonstration of analysis of big spatial data using R and Python.
Learning Objectives
Upon successfully completing this course, students will be able to:
- Identify the unique challenges of analyzing large-scale geospatial data.
- Gain proficiency in advanced statistical methods and AI techniques for geospatial data analysis.
- Apply Gaussian processes, graphical models, and scalable methods to real-world spatial datasets.
- Integrate modern machine learning techniques with traditional spatial models for comprehensive geospatial analysis.
- Implement efficient Bayesian inference methods suitable for big spatial data.
- Conduct a wide variety of geospatial analyses in R and Python
Methods of Assessment
This course is evaluated as follows:
- 30% Project proposal (midterm)
- 50% Final project and presentation (endterm)
- 20% Discussion