- Describe physical systems using the appropriate mathematical formulation. - Apply machine-learning algorithms to the resulting problem. - Understand the advantages and limitations of machine learning algorithms given the specific problem at hand.
Course Prerequisites
Introduction to physics as provided in the relevant first- and second-year classes. Basic knowledge of the Python programming language.
Teaching Methods
Each class will pair traditional lectures (to introduce the relevant problems) with hands-on exercises and demonstrations (to tackle the relevant problem). These computational activities are the key content of the course.
Assessment Methods
Students will develop a series of computational projects. These will be started during the lectures and completed asynchronously. The project report and associate codes, likely in the form of a Jupyter notebook, will then be submitted for evaluation.
Texts
- Statistics, Data Mining, and Machine Learning in Astronomy: A Practical Python Guide for the Analysis of Survey Data. Željko Ivezić, Andrew J. Connolly, Jacob T. VanderPlas, and Alexander Gray. Princeton University Press
Contents
- Probability theory. Bayes theorem. Descriptive statistics. - Bayesian vs frequentist statistics. From the pdf to the samples: inverse transform, acceptance/rejection. - Density estimation. From the samples to the pdf: histograms, Kernel Density Estimation. - Monte Carlo integrations. Markow chains. - Metropolis Hastings. MCMC diagnostics. Modern samplers. - Bayesian model selection. Savage-Dickey density ratio. - Computing the evidence. Nested sampling. Modern samples. - Project.
