| 1 |
Different methods have different trade offs
-
Robustness, accuracy, speed, suitability for parallelization,
-
problem size dependence
|
| 2 |
Neural networks do simple things on large data sets and parallelize easily
|
| 3 |
Expert systems do complex things on small data sets and parallelize with difficulty
|
| 4 |
Combinatorial Optimization
-
Finds exact minima in a time that is exponential in problem size. However in particular cases, e.g. TSP, very clever special techniques make this quite practical - solve exactly 104 ® 105 city problem if we can parallelize
|
| 5 |
Physical Optimization
-
Finds approximate minima in a time that is sometimes only linear log(linear) in system size.
-
Sometimes we only want approximate minima
|
