Sample of N points for hierarchical clustering

[daniprec]

Hierarchical clustering does not work for too many points (>1e5).
We need to take a sample of N power load points per appliance (N=1e4).
It is necessary to find a method that ensures the sample of N points contains at least one activation of each potential status. Otherwise, we risk taking a sample of points full of 0's (because many appliances are inactive >90% of the recording period).

[daniprec]

⚠️ Notice: This project is no longer actively maintained, and this issue won't be addressed. Feel free to continue the discussion, but be aware that there won't be further updates.