Probabilistic Programming and Bayesian Data Analysis (COGS516 - METU - Graduate Informatics Course)
This course will examine the basic principles of probabilistic programming and Bayesian modelling, for analysis of data which may come from observational or experimental cognitive science studies. A variety of Bayesian data analysis will be discussed and implemented in an expressive probabilistic programming language. Approaches for model building, model checking and model validation will be discussed following a Bayesian workflow.
Introduction to Probabilistic Programming; generative modelling, Bayesian inference, executing probabilistic programs; exact inference; rejection sampling; importance sampling; Markov Chain Monte Carlo (MCMC); efficient MCMC Techniques; deep Probabilistic Programming; Probabilistic Programming applications.
- Probability and Inference
- Regression
- Categories and Curves
- Causality: Confounders and Colliders
- Overfitting and model comparison
- Inference, Markov Chain Monte Carlo, Assessing Convergence
- Generalized Linear Models
- Multi-level Models
- Ordered categorical outcomes
- Gaussian Processes
- Review, Further Topics in PPL