Modern methods in statistical learning
Преподаватель:
Департамент анализа данных и искусственного интеллекта: Доцент
This subject is divided into two sections, one section per module. The first section investigates modern tools in statistical learning for regression and classification problems. We start by discussing bayesian linear regression and classification techniques, before moving to regression techniques beyond the linear model, including shrinkage methods, splines, smoothing splines and neural networks. Next, we discuss tree-based methods, bagging, boosting, and random forests. We then introduce validation techniques, such as the bootstrap and cross-validation. In the second part of this course, we present time-series models for sequential data, including Markov models, hidden Markov models, Wiener filtering, Kalman filtering. We also investigate sampling algorithms, such as rejection sampling, importance sampling, Markov chain Monte Carlo, Gibbs sampling. The focus of this course is on the mathematical derivations of the algorithms, and R language will be used to explain implementation of the techniques.
Программа дисциплины (PDF, 541 Кб)