Advanced Seminar Large-Scale Graph Processing and Graph Partitioning (IN2107, IN4435)
|Language of instruction||English|
|Position within curricula||See TUMonline|
Preliminary meeting: July, 13th 2023, 4:15 PM - 5:15 PM online https://tum-conf.zoom.us/j/63927749880?pwd=MWxURVRWbkR5ZWVqSDBYOGlJWHRXdz09 Graphs are a fundamental data structure commonly used to model relationships between data points, e.g., links between web pages, friendships between users in a social network, etc. In the past decade, many specialized distributed systems have emerged that are optimized for managing and processing graph-structured data. To analyze large graphs, such as web graphs or social networks, distributed graph processing systems are used, where several compute nodes execute a graph processing algorithm in a distributed fashion in parallel. As a pre-processing step, the graph is partitioned into several disjoint parts distributed across the compute nodes. In this seminar, we will study several large-scale (distributed) graph processing systems for static and dynamic graphs and the training of graph neural networks. However, the focus will be on the scaleable training of graph neural networks. Furthermore, we study streaming and in-memory graph partitioners. More information: https://docs.google.com/presentation/d/1VyDWynccZM_HZjUoYE_R1KYX7I0XtuGbvZqFGPV9xbM/edit?usp=sharing Preliminary meeting: July, 13th 2023, 4:15 PM - 5:15 PM online https://tum-conf.zoom.us/j/63927749880?pwd=MWxURVRWbkR5ZWVqSDBYOGlJWHRXdz09
Basic knowledge of distributed systems.
Teaching and learning methods
Modulkatalog: IN2107 - Presentations - Written report with figures (ACM proceedings style), to submit 2 weeks after the presentation
Grade is based on written report with figures (ACM proceedings style) (50%) and presentation (50%)