Note: This is the second talk in a pair, and should still mostly make
sense for audience members familiar with standard compaction.
The first talk is "Hierarchical (Co-)Compaction for DGGS Cells".

We motivate the talk around the problem of designing delivery zones of
restaurants for a food delivery service. Isochrones provide a good starting
point: we select the delivery zone to be all the DGGS cells reachable within
some travel time of an origin point.

When doing this for many restaurants, storage costs can become a concern.
To help with this, we might compact the cell sets, representing the same region
with fewer cells by letting parent cells logically stand in for their children.

However, compaction will only get us so far. Near the boundary of a region, a
set often holds most, but not all, of a parent's children, and compaction can't
progress further on those partial groups.
Thus, we might notice opportunities to add cells that
complete such groups, letting them collapse to a single ancestor and compact to
much smaller sets. If these cells are only slightly outside of the travel time
target, they may be worth adding for the compaction benefit. We give up some
small isochrone fidelity for a large benefit in compactibility.

We can formalize this idea as an optimization problem that trades off isochrone
fidelity against compactibility. Because compaction follows the DGGS's own
cell hierarchy, the problem has structure we can exploit: rather than
search over arbitrary cell sets, we can walk up the hierarchy to solve it
efficiently. We'll explain this algorithm and demonstrate a Python package (to be
published before the talk) that implements it.

We'll look at examples across multiple DGGSs like H3, S2, and A5, showing that
significant compression can be achieved with small cost to isochrone fidelity.
We'll note that this approach can extend to other problems and
objective functions, including weighting by population or profitability,
for example.

We'll end by showing that this approach can generalize to co-compaction for even
greater storage savings.