Course Information
- Course Name: Advanced Network Management (ANM)
- Topic: Time Series Intelligence in Cyberspace
- Language: English
- Course Type: Graduate
Course Instructor
Dan Pei
Associate Professor
Department of Computer Science and Technology
Email: peidan(at)tsinghua(dot)edu(dot)cn
Office: Room 1014, Ziqiang Building 1
Course Introduction
Our digital world is a data-driven ecosystem, a dynamic and interconnected network, and this course teaches Temporal Intelligence through Time Series Analysis. It has applications in key domains such as finance, IT operations, the industrial internet, cybersecurity, smart cities, smart manufacturing, smart healthcare, smart energy, and earth sciences. Time-dependent systems are modeled as an Internet of Time Series, consisting of interconnected and interdependent systems. Each state of a system is modeled as a variable whose value changes over time, forming a time series.
The temporal behavior of these systems is analyzed using probabilistic models and advanced time series techniques, where each variable represents a state and the interconnections reflect dependencies among systems.
This course covers various algorithms (both classical and deep learning-based) for time series analysis, including forecasting, anomaly detection, causal discovery and inference, segmentation, time series augmentation, etc from recent research papers in top AI conferences. Then, through case studies of time series tasks, the course further explores time series applications like AIOps (AI for IT Operations).
Finally, the course discusses the latest progress in advanced topics like time series foundation models. Students are expected to complete personal assignments, paper presentations, and a project related to time series algorithms in recent top conferences.
Class Time and Location
Time: Wednesday 9:50am-12:15pm
Location: Room 3304, Classroom Building 3
Assignments & Project
- Individual Assignment: Time Series Visualization and Anomaly Detection
- Individual Paper Reading: Select and share a paper from top-tier conferences on time series (Paper list will be provided)
- Group Project (1-3 students per group): Self-defined, as long as it is related to time series
Grading Policies
- Attendance: 10%
- Personal Assignments: 20%
- Paper Reading & Presentation: 20%
- Group Project: 50%
Syllabus:
| Lecture | Time | Topic | Slides | Project |
| 1 | Week 1 | Introduction | – | – |
| 2 | Week 2 | Univariate Time Series Feature Engineering | – | – |
| 3 | Week 3 | Pairwise Distance & Similarity Measurement | – | – |
| 4 | Week 4 | Clustering | – | – |
| 5 | Week 5 | Time Series Segmentation & Change Point Detection | – | Paper Presentation |
| 6 | Week 6-7 | Time Series Classification | – | Paper Presentation, HW1 Due |
| 7 | Week 8 | Time Series Forecasting | – | Paper Presentation |
| 8 | Week 9-10 | Time Series Anomaly Detection | – | Paper Presentation, HW2 Due |
| 9 | Week 12-13 | Causal Discovery & Inference | – | Paper Presentation |
| 10 | Week 14-15 | Advanced Topics (Fault Prediction & Foundation Models) | – | Paper Presentation |
| 11 | Week 16 | Project Presentation | Not Available | Group Project Presentation |
Other Information
《Practical time series analysis: Prediction with statistics and machine learning》, by Aileen Nielsen
《MIT 6.S191 Introduction to Deep Learning 》 with video and slides.
Course Assistant
Zhe Xie
Email: xiez22(at)mails(dot)tsinghua(dot)edu(dot)cn