V1 is mainly focused on beam search which is a greedy best first search. There are still many issues with it but I want to experiment with different algorithms
I used to do a greedy depth first search. A goal score should be set and the program will terminate as soon as one result is found. This is optimised for speed but routes are not as good.
Overall, the heuristic should be improved. This way even with a smaller beam size, routes should be better generally.
V1 is mainly focused on beam search which is a greedy best first search. There are still many issues with it but I want to experiment with different algorithms
I used to do a greedy depth first search. A goal score should be set and the program will terminate as soon as one result is found. This is optimised for speed but routes are not as good.
Overall, the heuristic should be improved. This way even with a smaller beam size, routes should be better generally.