Hyperbolic Embedding Performance Benchmarks for Real and Simulated Networks

1.1 GB of supplemental data from an ICLR 2026 paper comparing popular hyperbolic embedding methods. The package contains performance and quality results for algorithms like Bläsius et al. (ESA 2016) and Nickel and Kiela's Poincaré (NIPS 2017) and Lorentz (ICML 2018) embeddings on real-life hierarchies, networks, and simulated networks. Author Eryk Kopczynski released it under a CC-BY-4.0 license on figshare in May 2026.

Use Cases

• Benchmarking computation time of hyperbolic embedding algorithms based on the described performance comparison
• Evaluating embedding quality on hierarchical structures based on the described real-life hierarchies
• Comparing algorithm performance on synthetic network data based on the described simulated networks
• Reproducing the results from the ICLR 2026 paper 'Bridging ML and algorithms: comparison of hyperbolic embeddings'

Strengths

• Dataset size is 1.1 GB, indicating substantial supplementary material
• Explicitly compares performance of multiple named algorithms (Bläsius et al., Nickel and Kiela) from 2016-2018
• Includes results on both real-life and simulated networks as mentioned in the description

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

Eryk Kopczynski

Collection Method

Likely contains experimental results and generated embeddings from the comparative study described.

Freshness

Last updated 2026-05-09 11:29:52