报告人：Prof. Xiang Zhou, City University of Hong Kong
题目: A Mathematical Introduction to Machine Learning
地点：数学楼二楼报告厅
摘要：This series of introduction to machine learning is prepared for students in math major, particularly the computational math students who are interested in understanding the math behind machine learning algorithms and applications of modern machine learning methods to problems in sciences. The perspective is from the applied and computational math (approximation, optimization, stochastic process, optimal control, etc), in contrast to the statistics and computer sciences. But the students from application areas (computer vision, NLP, etc.) who are interested in learning math are also warmly welcomed. The motivation is not to enumerate and comment on the existing thousands of algorithms one by one, but to try to exploit a unified viewpoint of computational mathematics in the right balance of abstractness and specificity. The topics are selected based on both the underlying math structure and the practical importance, which cover from the classic work to the very recently published journal articles. This talk will not cover any programming practice such as Tensorflow. 

I:  Basics 
Review the basic concepts and fundamentals of machine learning, such as linear regression, logistic regression,  Bayesian learning, bias-variance trade-off,  etc. This is the standard material in most classic statistical learning books.
时间：2018年4月2日13:30-14:30 

II: Deep neural network 
This part focuses on artificial  neural network models, including the universal approximator, back-prop, the training algorithms such as stochastic gradient descent, the myth of generalization error,  etc.
时间：2018年4月2日14:45-15:45 

III: Reinforcement Learning 
The reinforcement learning has been successfully applied to Alpha Go and Alpha Go zero. This parts starts with a full introduction of Markov decision process — the math foundation of reinforcement learning, followed by  dynamic programming and  the approximate methods.

时间：2018年4月2日16:00-17:00