Random Graph Sampling

Cross-source consensus on Random Graph Sampling from 1 sources and 5 claims.

1 sources · 5 claims

How it works

Bernoulli-p random graphs include each unordered agent pair independently with probability p. — Graph-SND: Sparse Aggregation for Behavioral Diversity in Multi-Agent Reinforcement Learning
The Horvitz-Thompson estimator for Bernoulli sampling is unbiased for SND. — Graph-SND: Sparse Aggregation for Behavioral Diversity in Multi-Agent Reinforcement Learning
The Horvitz-Thompson estimator uses 1/p weights and population normalization so it remains unbiased even with an empty sampled edge set. — Graph-SND: Sparse Aggregation for Behavioral Diversity in Multi-Agent Reinforcement Learning
The normalized Graph-SND statistic is a uniform sample mean conditional on at least one sampled edge. — Graph-SND: Sparse Aggregation for Behavioral Diversity in Multi-Agent Reinforcement Learning
Random graph sampling supports claims about estimating full SND because omitted edges are treated as missing samples rather than excluded relationships. — Graph-SND: Sparse Aggregation for Behavioral Diversity in Multi-Agent Reinforcement Learning