HEURISTIC SEARCH: BEST-FIRST (A* SEARCH)
- Best-first search uses an evaluation function, h(N), which
indicates the relative goodness of pursuing that node. If we
combine this with the partial path score, we can define a
general evaluation function:
f(N) = g(N) + h(N)
which can be used to evaluate hypotheses as they evolve.
If we always pursue the best path according to this evaluation
function, what are the merits of this approach? What constraints
must be placed on this function to guarantee an optimal solution?
How would that solution compare to other search algorithms?
- A general algorithm for searching such spaces can be defined
as follows:
- A search algorithm is said to be admissible if
it can guarantee an optimal solution.
- One possible solution to the traveling salesman problem using
best-first search might look like this: