Is there a way to quantify how stable clusters are with different number of runs? We find that in our data, certain clusters are very stable, and appear regardless of how many itterations are performed. The appearance of others seem to vary, even if we do hundreds of itterations, suggesting that the algorithm does converge, which I suppose suggests the number of populations to detect needs to be tuned. Can you suggest any best practices for approaching these types of problems?
Is there a way to quantify how stable clusters are with different number of runs? We find that in our data, certain clusters are very stable, and appear regardless of how many itterations are performed. The appearance of others seem to vary, even if we do hundreds of itterations, suggesting that the algorithm does converge, which I suppose suggests the number of populations to detect needs to be tuned. Can you suggest any best practices for approaching these types of problems?