•First
steps:
–Implement
benchmark algorithms.
–Find a
simple, toy domain to solve mathematically
–Look
for flaws in current algorithms and try to improve them.
•Influential
work:
–Solve a fundamental math problem, especially one that gives rise to new algorithm.
–Show
importance of technique from other field.
–Create
a new problem.
–Build impressive system that demonstrates new ideas or potential effectiveness of existing ones.