Efficient SMC Inference for Kingman's Coalescent M30
Dilan Görür and Yee Whye Teh

Kingman's Coalescent: Prior over trees
T4 T3

T2

T2

T1

T1

T1

T1

coalescent time of a pair is sampled only once exploits the memoryless proper ty of the exponential distribution calculations are reused

new SMC for hierarchical clustering
improved more
t h

computational cost efficient samples

ighter variance igher effective sample size