Improved large-scale graph learning through ridge spectral sparsification

Name: Improved large-scale graph learning through ridge spectral sparsification
Uploaded: 2026-10-08T11:30:52.874018+00:00
Duration: 519 s
Description: The paper addresses slow graph learning methods by combining spectral sparsification with Laplacian learning. It proposes a new distributed algorithm, DiSRe, to efficiently create an accurate subgraph approximation of large graphs in O(n log³(n)) time. This allows complex tasks like Laplacian smoothing and semi-supervised learning to run much faster on single machines.

The paper addresses slow graph learning methods by combining spectral sparsification with Laplacian learning. It proposes a new distributed algorithm, DiSRe, to efficiently create an accurate subgraph approximation of large graphs in O(n log³(n)) time. This allows complex tasks like Laplacian smoothing and semi-supervised learning to run much faster on single machines.

Open the full episode