search

cjprybol / Mycelia

MIT License
4 stars 3 forks source link

Use degree >= 3 nodes for pre-indexing targets #147

Closed cjprybol closed 4 months ago

cjprybol commented 2 years ago

If we take starting nodes for traversing the graph at random, even if we take a weighted sample, could still accidentally grab a node that is off on a very low frequency or erroneous offshoot path.

By taking all nodes with >= 3, we always enable the option to take the better of the two or more adjacent paths

Possibly consider taking a weighted subset of these hub nodes, rather than taking all of them. But by taking all of them, we get a good way of indexing the graph completely

can explore by:

  1. weighted walk forward from A until we find a hub node or find B
  2. weighted walk forward from B until we find a hub node or find A
  3. use the pre-calculated shortest path from A to B
  4. can either stop there for a possibly, but not necessarily guaranteed, shortest path
  5. Can continue shortest path searches from A and B UP UNTIL they are equal to or longer than the precalculated route from hub to hub, then we'll know for sure which is the shortest
cjprybol commented 4 months ago

done in version being prepared for preprint, done for now