Bayesian Methods
- Course Code:
- MSCA 32014
Bayesian inference is a method of learning in which Bayes' theorem is used to combine the previous knowledge with the new evidence in the data to form an improved posterior knowledge. Another name for such methods is probabilistic inference. Probabilistic Bayesian models form the foundation of the most modern algorithms of Machine Learning and Artificial Intelligence.
The focus of this course is an introduction to Bayesian approach. Many methods learned by students in Statistical Analysis, Linear and Nonlinear Models, Data Mining and Machine Learning will be reviewed from the point of view of probabilistic inference. We will look at hierarchical, mixture, robust, and non-parametric Bayesian models and learn how to use them in practical applications. Content will include using probabilistic models to make business decisions under uncertainty, analyze causation in the data, use probabilistic inference to assess risk of black swan events, account for uncertainty in project management and other applications.
Students will learn necessary facts of probability theory, Bayesian reasoning, Markov chain Monte Carlo using JAGS, STAN and PyMC. The course contains large number of interactive demonstrations, workshops with examples through which the lecturer shares his own hands-on experience with the students.
Prerequisite:
- MSCA 31010: Linear and Nonlinear Models