Gradient Based and Gradient free Optimization

Hi,

I'm trying to do a comparaison between the different opitimization algorithm in terms of Computation Time.

The algo that am using:

PSO_NM,GA,Fmincon,Pattern_Search,Sumulated_Anealing

From what i knew before, i suppose that the Gradient free Optimization algorithm are far slower then the gradien based one

But by doing my simulation i'm seeing the that the gradient free are faster

i was wondering why? can you suggest some supposation

i'm restrecting the boundary the most possible.

and my objectif function is very sensible in a way that if i change one coefficient value by 0.0001, i'll get inf as a results (a lot of peaks i suppose)

Answer 1

[Matt J]

and my objectif function is very sensible in a way that if i change one coefficient value by 0.0001, i'll get inf as a results (a lot of peaks i suppose)

Gradient-based optimizers are designed for smooth functions (ones which have gradients). It sounds like your objective is not smooth at all!