Locate samples based on condition

[AlePalu]

Overview

Be able to locate samples in the domain based on a condition. Conditions might be, for instance,

• presence of missing data points
• high/low concentration of observations

Given a condition, we should obtain a subset of points, maybe clustered in subregions, on which we can perform different actions, for instance, randomly select a subsample.

Possible sub-problems

• Filter points based on condition (locate all not-nan points)
• Cluster filtered points in sub-regions (use proximity of the elements they belong to?)
• Randomly sample points on a filtered group
• Introduce a notion of proximity, i.e., given a point, locate the nearest point
• Sample points at random on the neighborood of a clustered region
• Allow to identify regions with highest/lowest points concentration