Course Name: Applied Machine Learning

Course Code: S2293177

Semester: 4

Credit: 3

Program: Electronics

Course Module: Compulsory

Responsible: Yan Xu

E-mail: xuyan@tju.edu.cn

Department:Tianjin International Engineering Institute, Tianjin University

Time Allocation(1 credit hour = 45 minutes)

Course Description

Since machine learning can only be understood through practice, by using the algorithms, the course is accompanied with assignments during which students test a variety of machine learning algorithm with real world data. The course uses the MLDEMOS TOOLBOX that entails a large variety of machine learning algorithms.

• Binary and multi-class classifiers: LDA, GMM with Bayes, SVM, Boosting,

• Pattern recognition and clustering,

• Non-linear Regression,

• Markov-Based Techniques for Time Series Analysis.

Prerequisite

Probability, Linear Systems, Optimization Methods.

Course Objectives

To introduce the basic principles of the design and analysis of modern digital communication systems. Topics include source coding, channel coding, baseband and pass band modulation techniques, receiver design, channel equalization, information theoretic techniques, block, convolution, and trellis coding techniques, multiuser communications and spread spectrum, multi-carrier techniques and FDM, and carrier and symbol synchronization. Applications to design of digital telephone modems, compact disks, and digital wireless communication systems are illustrated. The concepts are illustrated by a sequence of MATLAB exercises.

Course Syllabus

1. Introduction

   1. Basic concepts

2. Supervised learning

   1. Supervised learning setup. LMS

   2. Logistic regression. Perception. Exponential family

   3. Generative learning algorithms. Gaussian discriminate analysis. Naive Bays

   4. Support vector machines

   5. Model selection and feature selection

   6. Ensemble methods: bagging, boosting

   7. Evaluating and debugging learning algorithms

3. Learning theory

   1. Bias/variance tradeoff. Union and Cher off/Hoeffding bounds

   2. VC dimension. Worst case (online) learning

   3. Practical advice on how to use learning algorithms

4. Unsupervised learning

   1. Clustering. K-means

   2. EM. Mixture of Gaussians

   3. Factor analysis

   4. PCA (Principal components analysis)

   5. ICA (Independent components analysis)

5. Reinforcement learning and control

   1. MDPs. Bellman equations

   2. Value iteration and policy iteration

   3. Linear quadratic regulation (LQR). LQG

   4. Q-learning. Value function approximation

   5. Policy search. Reinforce. POMDPs

Textbooks & References

• Richard Duda, Peter Hart and David Stork.Pattern Classification(2nd ed). John Wiley & Sons, 2001.

• Tom Mitchell.Machine Learning. McGraw-Hill, 1997.

• Richard Sutton and Andrew Barto.Reinforcement Learning: An introduction. MIT Press, 1998.

• Trevor Hastie, Robert Tibshirani and Jerome Friedman.The Elements of Statistical Learning. Springer, 2009.

Grade Distribution

Attendance: 20% Final Exam: 80%

Capability Tasks

CT1: To understand basic science, and to have analytical ability and the ability to integrate related knowledge.

CT2: To apply relevant professional knowledge to the field of science and technology: understanding of the basic concepts and its connotation, application of different methods and concepts which have been learned, capability of judging the scope and limitations of such applications.

CT3: To grasp methodologies and engineering tools: identifying, utilizing and solving problems. Even if the students are not familiar with the content, they can turn to computer tools for systematic analysis.

CT4: To carry out experiments in research environment with the abilities to utilize tools, especially for data collection and processing.

CT10: To have the capacity to work in international environment; the capability to master one or more foreign languages and be open to foreign cultures; be able to acclimatize themselves to the international language environment.

Achievements

• To choose an appropriate ML method for a given problem. -Level: M

• To assess / evaluate appropriately and comparatively ML methods given a set of data. -Level: A

• To appropriately apply ML methods. - Level: A

Students: Electronics, Year 2