Discussion of article "Population optimization algorithms: Saplings Sowing and Growing up (SSG)" - page 11

New comment
 
Maxim Dmitrievsky #:
Sort of cross validation selects the best needle or surface. And to get many needles, you can optimise across different pieces of history. The ph-i remains the same.

Apparently, this is about some other needles than I said earlier.

 
fxsaber #:

Apparently it's about some different needles than I was talking about earlier.

About the mapping ones. Split the sample into 10 chunks, on each optimise and discard sets with unique annoying peaks from each chunk that are not in the others. Then, if you exclude them from the overall optimisation over all the data, it will be peace and quiet and God's goodness. But it's not exact, just made it up. I don't know how to exclude arbitrary ranges in opt variables.
 

Maxim Dmitrievsky #:
Про картиночные. Разбить выборку на 10 частей, на каждой оптимизировать и выбросить сеты с уникальными раздражающими пиками из каждого куска, которых нет в других.

In terms of computational resources, it is equal to any 10 optimisations.

Then, if you exclude them from the overall optimisation on all data, it will be peace and quiet and God's grace. But it's not exact, I just made it up.

And we find one hill.

I don't know how to exclude arbitrary ranges in opt variables.

Okay.

 
fxsaber #:

Computationally, this is equal to any 10 optimisations.

And we find one hill.

Okay.

The end goal is to find the hills, the needles are intermediate hills.

Right, well that's pretty much the default in the MoD. If the data is rubbish, it usually doesn't do much. If it's not rubbish, it's not much needed :)
 
Maxim Dmitrievsky #:
The end goal seems to be to find the hills

More than one. That's why I suggested the sequential discard option.

 
Andrey Dik #:

there's a lot of algorithms, I don't know if there are even cooler algorithms.

The table is live, I add algorithms to them as I learn them, i.e. I can't say - that one over there is the coolest, I only know the ones I described))))

In fact, you could already take ant, bee and weed, they are very good. wooden of course now tears everyone, what will be the next leader - I do not know.

I will get to the hybrid ones when I will go through all the significant known ones, hybrid ones are very promising.

For now I am considering population types, but there are other types, it will be interesting to study them too.

maybe there is one already :)

 
Maxim Dmitrievsky #:

maybe there's one already :)

Yes, very interesting organism.))))

but, the slug uses only 2 dimensions, and even the simplest AOs can cope with 2 dimensions. how it will behave in 1000 dimensions is a big question, because the complexity of the problem grows non-linearly with the number of dimensions.

 
Andrey Dik #:

as the number of measurements increases, the complexity of the problem grows nonlinearly.

For self-education, what is the dependence of complexity on measurements?

 
fxsaber #:

For self-education, what is the dependence of complexity on measurement?

I confess I don't know. I only know that it grows non-linearly fast.

Aleksey Nikolavev appeared here, maybe he knows the exact answer to this question. I forgot which way to call a forum user.


I am currently checking an article about electromagnetic search - EM, with mediocre, in general, characteristics, it has one property that struck me.

 
fxsaber #:

For self-education, what is the dependence of complexity on measurement?

Exponential, in the general case
1...45678910111213
New comment