Map Partition
Given some distribution over 2d space, can we partition it in a way that the partition is intuitive for users to view (one region per partition index) and the partition regions have some desired distribution?
This solution is inspired by the following analogy. Consider filling a landscape of varying height (probability distribution on left) by 6 large water balloons (partitioning on the right).
This is a bit abstract ... but we intend to use this partitioning to select a point from a 100 x 100 array by querying a user who can only answer with 6 responses. We will apply this inference system to a destination based wheelchair control using SSVEP BCI.