The Elements of Statistical Learning

1. Basic Course Information

Course Title: The Element of Statistical Learning
Course Code:102374
Credit Hours/Credits:64/4
Students: Undergraduate Students majoring in Applied Statistics 
Prerequisites: Mathematical Analysis,Advanced Algebra,Probability Theory and Mathematical Statistics.
Subsequent Course(s):None
Course Category: Professional Foundamental Course
Assessment: Final exam(70%) + Participation and Assignment(30%)

2. Course Introduction 

Statistical learning plays a key role in many areas of science, finance, economics and industry. This course is about learning from data and is also a compulsory and important subject in Applied Statistics. Three subjects are covered in this course: supervised learning, unsupervised learning and reinforcement learning. The very first model discussed in supervised learning is linear methods for regression and classification. After introducing the concepts of bias and variance, the second topic touches model assessment and selection. Finally, it discusses a bunch of methods such as regularization/penalization methods, kernel methods and local weighted regression for battle against over fittings.

In unsupervised learning, because of the absence of supervision signal in data, several methods such as the k-means clustering algorithm, EM algorithms, factor analysis models as well as principal components analysis and independent components analysis, are covered to solve the task that is rather to describe how the data are organized or clustered. In the reinforcement learning framework, only a reward, instead of unambiguous “right answer” is given in training data, the last but not least algorithm in this part is Markov decision processes(MDP) which can figure out how to choose actions over time so as to obtain large rewards.