We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: ${\rm N{\small ode}S{\small ig}}$: Random Walk Diffusion meets Hashing for Scalable Graph Embeddings

Abstract: Learning node representations is a crucial task with a plethora of interdisciplinary applications. Nevertheless, as the size of the networks increases, most widely used models face computational challenges to scale to large networks. While there is a recent effort towards designing algorithms that solely deal with scalability issues, most of them behave poorly in terms of accuracy on downstream tasks. In this paper, we aim at studying models that balance the trade-off between efficiency and accuracy. In particular, we propose ${\rm N{\small ode}S{\small ig}}$, a scalable embedding model that computes binary node representations. ${\rm N{\small ode}S{\small ig}}$ exploits random walk diffusion probabilities via stable random projection hashing, towards efficiently computing embeddings in the Hamming space. Our extensive experimental evaluation on various graphs has demonstrated that the proposed model achieves a good balance between accuracy and efficiency compared to well-known baseline models on two downstream tasks.
Subjects: Machine Learning (cs.LG); Social and Information Networks (cs.SI); Machine Learning (stat.ML)
Cite as: arXiv:2010.00261 [cs.LG]
  (or arXiv:2010.00261v1 [cs.LG] for this version)

Submission history

From: Abdulkadir Celikkanat [view email]
[v1] Thu, 1 Oct 2020 09:07:37 GMT (1203kb,D)

Link back to: arXiv, form interface, contact.