-
Completeness: Will the algorithm eventually find a solution if one exists?
- depends on branching factor, cycles, and termination conditions
-
Optimality: Will the algorithm return the best solution according to the cost function?
- requires admissible evaluation or exhaustive guarantees
- often sacrificed in exchange for speed or scalability
-
Time complexity: How many nodes are generated/expanded?
- Typically exponential in depth for uninformed search
- Strongly affected by branching factor and heuristic guidance
-
Space complexity: How much memory is retained during search?
- Includes frontier/fringe storage and closed lists
- In practice, this is often the first limiting factor
- Motivates memory-bounded and local search methods