Course Syllabus for

EST 5010 Machine Learning (Fall 2021)


Catalog Description

This course offers you an introduction to pattern recognition and machine learning and focuses on deep learning and neural networks. We cover fundamental concepts and various neural network models, how they are trained and tested, and how they can be deployed in applications in computer vision, natural language processing and robotics, etc.

Prerequisite

Knowledge of linear algebra, calculus and elementary probability is necessary.  A little exposure to optimization and machine learning will be helpful.

Course Goals

· To understand fundamental methods of pattern recognition, machine learning, deep learning and deep neural networks (DNNs);
· To be able to build, train and apply supervised and unsupervised deep learning with fully connected, convolutional, generative, recurrent and reinforcement DNNs;
· To understand key parameters in a DNN architecture;
· To apply deep learning algorithms to solve computer vision, natural language processing, robotics, and other data analytics problems.

Topics


Week Topics Lecturer
1

 Introduction

·AI and Machine Learning
·Data Science Fundamentals

Stan Z. Li
2

 Multi-Layer Perception

·Nonlinear Regression
·Back Propagation and SGD

Yue Zhang
3

  Loss Functions

·Minimum Squares, Bayesian, Information Theoretic Loses
·Regularization Methods

Yue Zhang
4

 Convolutional Neural Networks

·Convolution Operations
·AlexNet, VGG, ResNet, NIN, Inception

Stan Z. Li
5

 Graph Neural Networks

·Graph vs Vector Neural Networks
·Graph Convolutional Networks

Stan Z. Li
6

 Generative Autoencoder Networks

·Autoencoders
·Variational Autoencoders

Stan Z. Li
7

 Recurrent and Recursive Neural Networks

·Hidden Markov Models, Conditional Random Fields
·RNN/LSTM,Transformer/BERT

Yue Zhang
8

 Sequence Search and Parsing

·Tree Structures and Conditional Random Fields
·Perception with Inexact Search,Transition-based Parsing.

Yue Zhang
9

 Transfer Learning and Domain Adaptation

·Inductive and Transductive Transfer Learning
·Unsupervised Transfer Learning

Donglin Wang
10

 Meta Imitation Learning

·One-Shot Learning, Model-Agnostic Meta-Learning
·Conjugated Task Graph

Donglin Wang
11

 Generative Adversarial Networks

·Generative vs. Discriminative Algorithms
·Basic GAN, Conditional GAN, CycleGAN, Adversarial AutoEncoder

Donglin Wang
12

 Reinforcement Learning

·Q-Learning, Policy Gradient
·Deep Q Network, Actor-Critic Algorithm

Donglin Wang
13

 Application: Computer Vision

·Person Re-Identification

Zhenzhong Lan
14

 Application: Natural Language Processing

·Building a Chatbot

Zhenzhong Lan
15

 Application: Robot Learning

·Visual SLAM

Zhenzhong Lan
16

 Project Report Presentations

All

Instructors

·Dr. Stan Z. Li Stan.ZQ.Li@westlake.edu.cn
·Dr. Yue Zhang ZhangYue@westlake.edu.cn
·Dr. Donglin Wang WangDongLin@westlake.edu.cn
·Dr. Zhenzhong Lan LanZhenZhong@westlake.edu.cn

Course Project and Evaluation

Each student should do a course project, preferably in connection with his/her own research, under the guidance of one of the four lecturers. The students are advised to start thinking about the project from the beginning of the course.

Evaluation Item Score Submission Deadline
 Project Proposal 30% Week 10
 Project Report 60% Week 16
 Lecture Attendance 10%
 A: 90 and above / B: 80-89 / C: 70-79 / D 60-69 / F<60 .

Course Tipis

·Getting a good understanding of fundamental knowledge in mathematics, pattern recognition and traditional machine learning, as covered in part I of textbook[1], is strongly recommended.
·Regular attendance is highly recommended.
·You can find information about online resources here.



Westlake University
Shilongshan ST #18, Xihu District, Hangzhou, Zhejiang Province, CN
中国浙江省杭州市西湖区云栖小镇石龙山街18号
2020 BY Westlake Machine Learning  浙ICP备16029590号