in reply to A different kind of search algorithm?
That sounded like simulated annealing and what I've read recently says there isn't any generally useful preset values for your search algorithm. You could consider the problem of getting optimal preset values as just another search problem. If you had no useful ideas for initial values, you can start with random values and search for good search parameters.
[Added: I briefly wondered "Was this the A* algorithm?" but I was wrong. I could likely find it in a nearby textbook but um... it's late and I'm more motivated to go to sleep.]
⠤⠤ ⠙⠊⠕⠞⠁⠇⠑⠧⠊
|
|---|