What’s wrong with what exists already?
They reflect limited data volumes, slow computing time, poor performance, poor flexibility of the choice of objective functions, etc
Conventional heuristic oriented algorithms
- Only developed for searching local optimum
- Only capable of small and simple data testing
Conventional optimisation oriented algorithms
- Hill climbing search manner (easy to get stuck at local optimum)
- Huge computational times to converge (near) global optimum if possible
- High level of mathematical concepts and theoretical sophistication
- Limited operational performances for large data