Evaluating Algorithms - Correctness
Correctness or quality of the answer
- Does it give the right answer.
- Does it give an answer that is close to the right answer (for an approximate algorithm).
- This can be extremely difficult to determine.
- Does it give a good answer on real data or on what I foresee as real data.
- Must implement and test on real data.
- Use of benchmarks
- Good because common to all.
- Bad because algorithms can be tuned to a benchmark.