Handbook

Graphs are ubiquitous and are widely used to capture relationships between different entities in real-world applications. However, compared with traditional sequential data (e.g. text and audio), the unstructured property and the sparsity make processing big graphs very challenging. The course will introduce a series of data structures and algorithms for graph processing in terms of database (i.e., big data) and deep learning. Fundamental methods and the state-of-the-art research works will be integrated. The course will serve as a launching pad for those interested in graph analytics, big data processing and graph neural networks.

Data structures and algorithms are the building blocks of many complex systems and software. Certain fundamental graph algorithms such as Dijkstra's algorithm and depth-first search have been covered by many text books and compulsory courses. They may be discussed in terms of pseudocode and time complexity. This course will start from studying how to efficiently implement the fundamental algorithms in big graphs. Then, the course explores more challenging and more complex algorithms step-by-step. When dealing with big graphs, we may consider various scenarios such as external memory solutions, distributed solutions, multi-core solutions, etc.

The course also puts some attention to graph neural networks, which is a hotspot in the area of AI and deep learning. The course will not study theoretical details about machine learning and deep learning but just introduce several representative graph neural networks. The students will play with basic graph learning tasks and understand learning-based techniques for graph problems such as link prediction and node classification.