Statistics and Machine Learning for the Physical Sciences

Catalog description

In this course, we will learn about new methodological research at the intersection of statistics, machine learning, and AI with applications in the physical sciences (including meteorology, remote sensing, astronomy, and particle physics). Students will see examples of how core statistical ideas (confidence sets, hypothesis testing, two-sample testing, and uncertainty quantification) play a key role in cutting-edge research in predictive inference, probabilistic forecasting, anomaly or signal detection, and likelihood-free (aka simulator-based) inference. Homework assignments will include assigned reading, theory problems, and some simple computer exercises. Students will work with real-world data during TA-led computer labs using Python/PyTorch and Jupyter notebooks to apply the skills and knowledge acquired throughout the program. The course assumes a solid foundation in the theory of probability and statistical inference at the level of 36-226 . Prerequisites: 36-218 or 36-236 or 15-259 or 36-226 or 36-326