Massively Parallel Algorithms for High-Dimensional Euclidean Minimum Spanning Tree

Explore a cutting-edge lecture on Massively Parallel Algorithms for High-Dimensional Euclidean Minimum Spanning Tree problems. Delve into the latest advancements in distributed algorithms for clustering large-scale, high-dimensional datasets. Learn about the challenges of solving Euclidean Minimum Spanning Tree (MST) problems in the Massively Parallel Computation (MPC) model and discover a novel approach that achieves a constant factor approximation in O~(loglogn) rounds. Understand the limitations of previous tree-embedding methods and how the presented algorithm combines graph-based distributed MST algorithms with geometric space partitions to overcome these constraints. Gain insights into the application of this technique to the Euclidean Traveling Salesman Problem (TSP), achieving a significant improvement in round complexity. This talk, presented by Peilin Zhong from Google, offers valuable knowledge for researchers and practitioners working with massive transformer-based embeddings and other high-dimensional data clustering challenges.

Massively Parallel Algorithms for High-Dimensional Euclidean Minimum Spanning Tree