Algorithm of Maximization
The maximization process is based on the variables marker classes represented in a core collection.
For each core collection or subcore collection, a score is calculated on marker variables as the number of represented classes.
To do it as well, Axij matrix is used by counting values above or equals to 1 for the core collection accessions
- 1st step : random choice of one core collection of n accessions from data table N accessions
- 2nd step : creation of N subcore collections by droping each accession, one by one.
- 3th step : a score is calculated for each subcore.
- 4th step : we drop one accession from the core collection where the subcore score is the highest because it's the accession which gives the less diversity. If 2 accessions are equals, a 2nd diversity criterion of these both two subcores is calculated. The accession which has the less one is dropped out.
- 5th step : we set up core collections of N accessions by adding each accession which is not already in the subcore
- 6th step : the score of each core collection is calculated. We add the accession which builds the core collection having the highest score on active marked variables
While the core collection score is improved, we repeat each step.
To avoid perpetual loop (when the score is first incremented then decremented by 1 in each iteration), we define a breaking value for the loop iterations.
Note: AxijFreq is used by optimizing functions for Nei and Shannon indices.
<
>
^