Numerical series density - page 8

 
Vyacheslav Kornev:

you won't find the densest points of any row more accurately

that's the only way to find the point that's closest to all the others - i.e. where the density is.

Am I good?)

Well done for thinking about this topic.

However, your method will only find one number without assigning it to a group of numbers representing a cluster. Imagine a starry sky - there are areas on it with a dense scattering of stars and there are single stars, our task is to find an algorithm to determine these densities.



 

No, the first picture has two numbers.

And the second

you, you can put the numbers in ascending order of calculated density, and take as many as you want

 
-Aleks-:

Well done for thinking about this topic.

However, your method will only allow us to find one number, without assigning it to a group of numbers representing a cluster. Imagine a starry sky - there are areas in it with a dense scattering of stars and there are single stars, our task is to find an algorithm to determine these densities.



so please

find the densest numbers as I've written.

If you need any clusters.

then enter what parameters you need and what to take as a cluster.

let's say you consider as a cluster a number next to which there are 3 or more numbers, at a delta distance of no more than 5

I counted all the deltas for the numbers, saw who has 3 or more numbers with a delta less than 5.

and got a series of numbers

 

with my method you can

1. Find the densest numbers/numbers in the whole number series

2. Find clusters

3. again find specifically in these clusters the densest points

 
Vyacheslav Kornev:

No, the first picture has two numbers.

And the second

you, you can put the numbers in ascending order of calculated density, and take as many as you want.

So the thing is, you don't know initially how many you need. So how much is a relative question...

Let's take the numbers I suggested earlier and use your algorithm

Now let's arrange the numbers sequentially - depending on the sum of deltas, and compare it with my proposed algorithm

Sum of deltasAlgorithm 1Algorithm 2
15471310
16705613
16706515
16805121
16827140
17004642
17107846
17244251
17288156
17404078
18303181
195021
203415
212010
2600190
2930223
3038232
3290250
3450260
8580545

Numbers from 190 can be discarded, but what about 65, 71, 31? Also, as I noted earlier, it is not clear how to group these numbers together.


 
Vyacheslav Kornev:

with my method you can

1. Find the densest numbers/numbers in the whole number series

2. Find clusters

3. again find specifically in these clusters the densest points

Demonstrate on the numbers above - point by point - maybe I don't fully understand the method. Provided we don't know the density parameters beforehand - i.e. we can't fit them to a particular numerical series.

 

that we do not know the parameters of the density of the pre

That can't be right.

If you don't know initially what to take as density, at least starting from which value the density starts you won't find it

 
I don't understand the picture above the first one
 
give me a small series of numbers up to 10 numbers I will show you
 
Vyacheslav Kornev:

that we do not know the parameters of the density of the pre

That can't be right.

If you don't know initially what to take as the density, at least starting from which value the density starts you won't find it

Why can't it do that? There are rows of numbers and you have to make a decision... My algorithm (I haven't tested it much) is able to do this.

Vyacheslav Kornev:
I do not understand the picture above first

Table from excel did not work to insert - limit on the number of characters in the forum. The table shows your method, for each number (second column) is delta (made modulo) relative to all other numbers, then the delta is summed up.

Vyacheslav Kornev:
give me a small series of numbers up to 10 numbers I will show

10 is not enough - take the above numbers for clarity.