Discussion of article "Population optimization algorithms: Saplings Sowing and Growing up (SSG)" - page 5

New comment
 
fxsaber #:

ZY It is strange that "smart" people with suggestions of smoothing do not understand the nature of the formation of the surface object.

I formulated the question very badly, it is strange that I was not minuted at all
 
mytarmailS #:
Well, then it's quite simple, as I wrote above.
You need 20 peaks, just run the AO 20 times.

You'll end up with an opt file with 20 runs of optimisation. Where are these 20 peaks in it?

 
fxsaber #:

So you get an opt-file with 20 optimisation runs. Where are these 20 peaks in it?

Well, the result of the optimisation
i.e. the best solution found.
You mean the parameters you were looking for.

This is the peak on the multidimensional surface of all possible parameter variants.
 
mytarmailS #:
I phrased the question very poorly, I'm surprised I didn't get banned.

The wording is exhaustive. The smart ones were either too lazy or didn't see the question.

 
mytarmailS #:
Well, the result of the optimisation
i.e. the best solution found
So the parameters you were looking for

This is the peak on the multidimensional surface of all possible variants of parameters.

This is only ONE peak.

 
ah that's what a hedgehog needs.....
well, then bee, cuckoo, and monkey and bacterial. these algorithms cluster in all peaks practically (if possible, if the population size is commensurate with the number of peaks).
 
fxsaber #:

That's only ONE pic.

Well, one full AO run == one peak.

20 runs == 20 peaks.

Or am I missing the point?
 
For such specific tasks, one can think of a "kick-out" mechanism, when a group that has grown too large is kicked out of the group, which is forced to form clumps in separate extremes.
 
mytarmailS #:
Well, one full AO run == one peak.

20 launches == 20 peaks.

Or am I still missing the point?

a good alg will find the same peak, what's the point? or deliberately use a crappy alg?
 
Andrey Dik #:

a good alg will find the same peak, what's the point? or deliberately apply a crappy alg?
Limit the number Iterations.
Randomised initial parameters

If the space is large, it will almost never find the same thing.

It is also possible to penalise in FF for similarity of current parameters with past ones already found.

So, everything is solvable.

12345678910111213
New comment