Here few separate problems needs to be discussed and taken into account:
The first problem is (as far as I understand) that in described situation problem is that 'springs' between considered points (data entries - rows) are already extremely contracted and flattening/changing space there is like division by 0 in those regions of space
solution to that problem might be to abstract and create abstract point where multiple points from original data will be represented as one , and after algorithm processing we may recover and assign final solution to original data as if original point are in the same point in result
such solution is altering sammon mapping alghoirtm
Completly another problem is basic assumption (for dendryt part) that no two distances can be the same. Used algorithm must create a tree, tree that uses particular set of edges . If two eggs are same in length/cost there might be the case where we choose different set of edges and still get valid output a feasible solution of dendryt-wrocławski. Such behavior must result in one of :
The result is non deterministic (each time we may get differen result)
(already applied) The data must not contain confusing edges - same in length where we cannot distinguish choosing one is better than another: The set of points merged together in single point (1.) definitely falls into this schema, there are multiple connections between them that share common length (a 0 length) we are confused which to use .
3 We may delete data based on certain criteria (this is dangerous )
4 maybe we can modify idea of dendryt-wrocławski alghoritm
Here few separate problems needs to be discussed and taken into account:
The first problem is (as far as I understand) that in described situation problem is that 'springs' between considered points (data entries - rows) are already extremely contracted and flattening/changing space there is like division by 0 in those regions of space
Completly another problem is basic assumption (for dendryt part) that no two distances can be the same. Used algorithm must create a tree, tree that uses particular set of edges . If two eggs are same in length/cost there might be the case where we choose different set of edges and still get valid output a feasible solution of dendryt-wrocławski. Such behavior must result in one of :
3 We may delete data based on certain criteria (this is dangerous ) 4 maybe we can modify idea of dendryt-wrocławski alghoritm