TODO: Random distributed HDA*

jinnaiyuu / Hash-Distributed-Astar

Hash Distributed A*

https://sites.google.com/site/yuujinnaishomepage/home/parallel-search

MIT License

3 stars 1 forks source link

TODO: Random distributed HDA* #34

Closed jinnaiyuu closed 9 years ago

jinnaiyuu commented 9 years ago

Check the fairness of Zobrist hashing.

If the random generator is completely random, then the output is totally even.

Implement random function as a distribution hash function. If the random is mathematically fair, the "bind" would be a one single line. If not, then it will make a bind with a width.

jinnaiyuu commented 9 years ago

Implemented. Experiment going on.

jinnaiyuu commented 9 years ago

Done. Seems:

Tons of duplicates.
Would NOT deliver nodes equally.

Why does it have a unevenness?

jinnaiyuu commented 9 years ago

Seems it came from the unevenness of work itself. Some node is just a duplicate so that it does not need to be expanded. Some nodes generate three nodes.

Even with a cyclic work distribution, there still exist a unevenness.