Discussion of article "Population optimization algorithms: Saplings Sowing and Growing up (SSG)" - page 3

New comment
 
Anatoli Kazharski #:

You need a mode that finds all hills and gives these ranges for all parameters that you can work with further.

I am afraid that heuristics will not allow this.

Subsequent optimisations can be performed only in the ranges of such robustness hills.

That is why it will most likely not work that way.


But cutting out the region of global maximum and running GA without it is quite possible. At each such iteration we will have a global maximum without taking into account the previously found ones.

 

Suppose there are optimisation results for all combinations of parameters over a long period of history.

There is a graphical interface with a time scale, moving a slider on which you can see all the robustness hills and how they float over time.

This would be a very useful tool.

 
Anatoli Kazharski optimisation results for all combinations of parameters over a large span of history.

There is a graphical interface with a timeline, moving a slider on which you can see all the robustness hills and how they float over time.

This would be a very useful tool.

Either I'm not getting it right, or I'm seeing too huge a number of calculations.

 
it looks like a separate topic for an article.
It is necessary to understand the concept of FF.

 
fxsaber #:

Either I'm not getting it right, or I'm seeing too huge a number of calculations.

Ideally, of course, you should have all results after full optimisation. But it may not be necessary.

I also need a tool to visualise the ranges of robustness hills on the scale of all parameter ranges.

Let's say (roughly) that these could be the ranges that performed better in combination with other ranges (another colour shows the second range from another hill):


 
Andrey Dik #:
requires a breakdown of the concept of FF.

The abbreviation is not understood.

Andrey Dik #:
it seems that a separate topic for an article is coming up.

First, you can try to select the area around the found global GA with the standard GA, because the opt-format of the optimisation results is fully open.

And it would be interesting to add the regular GA to the comparison table.

AO

Description

Rastrigin

Rastrigin final

Forest

Forest final

Megacity (discrete)

Megacity final

Final result

10 params (5 F)

50 params (25 F)

1000 params (500 F)

10 params (5 F)

50 params (25 F)

1000 params (500 F)

10 params (5 F)

50 params (25 F)

1000 params (500 F)



 
Anatoli Kazharski #:

Ideally, of course, you should have all results after full optimisation. But it may not be necessary.

A tool to visualise the ranges of robustness hills on the scale of all parameter ranges is also needed.

Let's say (roughly) that these could be the ranges that performed better in combination with other ranges (another colour shows the second range from another hill):

Apparently continuing to completely misunderstand the idea.

 
fxsaber #:

1.The abbreviation is not understood.

2. At first you can try to select the area around the found global by standard GA, because the opt-format of the optimisation results is fully open.

And it would be interesting to add the regular GA to the comparison table.


1. fitness function, adaptability
2. This may not be necessary in the search for robust parameters. that is why the idea has arisen, or rather has been in my mind for a long time. even have a draft article about 6 years old.
3. it is a complex and delicate issue. there are testing results, but there are a lot of buts. if I am allowed, I am ready to share in the table in the next article.
 
Andrey Dik #:
1. fitness function, adaptability

I am weak in terminology. If it is an optimisation criterion, I don't see why it is needed for this problem. If a test subject is needed, then Forest is fine.

 
fxsaber #:

I am weak in terminology. If it is an optimisation criterion, I don't understand why it is needed for this task. If you need a test subject, then Forest is fine.


Forest is a FF. i.e., it is some requirements that have been imposed on the system being optimised. if the requirements to the system are changed, the FF will change, but the system has not changed, right?
It's as if the user tried to change the requirements in various ways and still got Forest. Integral FFs look exactly like Forest, such as balance, for example.
It is necessary to try not to use integral FFs, if possible, and if it is not possible, then to make NADstroika over FFs, i.e. to apply FFs to FFs in order to avoid sharp peaks. any classtering over balance FFs is NADstroika over FFs.
Well, let's take an example. we have taken a balance FF. it looks (presumably) like Forest. you can fish in the murky water of the results of optimisation by balance for particles of sense that lie somewhere near, and you can go the other way, you can superstructure the balance FF so that the surface is no longer acutely finite, and all the necessary parameters lie near and at the same time on the topmost gentle hill!
In short, we can say that if the FF is acutely finite, then either this is really all that can be squeezed out of the problem, or the researcher has made a mistake.
12345678910...13
New comment