Reproduce “Local Intrinsic Dimensionality for Dynamic Graph Embeddings” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Dušica Knežević, Miloš Savić, Miloš Radovanović

Researchers from the University of Novi Sad adapted the NC-LID measure for dynamic networks to assess its relationship with dynamic graph embedding quality. They found a consistent negative correlation between NC-LID values and embedding quality metrics, indicating that nodes with higher local intrinsic dimensionality have weaker embedding representations.

Key insights

  • NC-LID, adapted for discrete-time dynamic networks, is a relevant indicator for assessing the intrinsic quality of dynamic graph embeddings.
  • High NC-LID values consistently correlate with "weak" embedding representations, specifically for nodes whose vectors poorly preserve temporal local connectedness.
  • Incorporating LID-awareness into dynamic graph embedding algorithms could improve their ability to represent complex evolving network structures.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.