It may be advantageous to search forward (from starting state towards the goal) or backwards (from goal to the starting state). Criteria include:

• Number of initial and goal states.

• Direction of greatest branching factor.

In many AI problems, backward reasoning is strongly preferred. The reason is that starting from the goal causes many variables to be bound to constants, greatly reducing the branching factor in the backward direction.

Bidirectional search, searching in both directions until the fringes of the searches meet in the middle, can save time: 2 * b^d/2 < < b^d . However, this may take a lot of storage.

Contents    Page-10    Prev    Next    Page+10    Index