1 of 3

Heuristic Sampling�(Chung-Lu)��Fast, but Biased ☹️

The “Watt Engine” for the Config. Model�unbiased, 100x faster random network sampling

Exact Sampling�(Max-Ent)

�Unbiased, but Slow ☹️

Our Method�(MaxEnt)��Unbiased ✅ �100x Faster 🚀!

2 of 3

Brute-Force Sampling

Proposed

Speed via Skipping: Sample candidates, then rejection sampling – we do not compromise accuracy 😉

Edge Sampled

Evaluated

Last evaluated

Evaluation Skipped

Adjacency matrix

3 of 3

📧 skojaku@binghamton.edu

Xuanchi Li, Ph.D student

600k nodes

Median Runtime (s)

230s

57h

Density

Clustering coefficient

MaxEnt

Chung-Lu

Empirical

Enable accurate significance assessment for large networks�(e.g., motif analysis; working on it 😉)