Algorithm 2: Poisson-Gamma downsampling
Input: $N\times M$ data matrix X; parameters $s,r>0$;
1. Draw $Z_1,\dots ,Z_M{\sim }^{i.i.d.}\Gamma \left(s,r\right)$.
2. Draw ${\tilde{X}}_{ij}\sim Poisson\left(X_{ij}{Z}_j\right)$, for $i=1,\dots ,N$ and $j=1,\dots ,M$.
Output: $N\times M$ downsampled data matrix $\tilde{X}$.