You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As of now, the HclustParam class, has 2 slots (metric and dist.fun) that are distance related. These slots are used in
The show method.
The clusterRows method, where the slots are used to call dist.fun with the metric parameter.
The idea of this abstract class would be to have these two slots so that none of the classes that inherits from it would have to declare them. This class would also handle the show method so that the sub-classes don't have to. It could have a constructor that handles the slots (if a slot is NULL: call standard `dist``). Finally, this class could have a method that gives the distance matrix.
Here is a little diagram to illustrate the concept.
The drawback of this idea is that the inheriting classes need to use the distance matrices. And I haven't found another class in bluster that uses it. At start, I thought this could help improve the maintainability of the code, but if we can't find other clustering method that use this, it may be of no use.
So as a sum up this class would be useful to any class that uses a distance matrix. The idea would be that these classes would inherit from the abstract class but maybe there are other possibilities.
What do you think?
The text was updated successfully, but these errors were encountered:
As you said, the idea is good if there is more than one method utilizing this general class. Otherwise, I think this just adds little bit more complexity. --> we can leave this issue open, and implement it when we find use-cases.
As of now, the
HclustParam
class, has 2 slots (metric
anddist.fun
) that are distance related. These slots are used inshow
method.clusterRows
method, where the slots are used to calldist.fun
with themetric
parameter.The idea of this abstract class would be to have these two slots so that none of the classes that inherits from it would have to declare them. This class would also handle the
show
method so that the sub-classes don't have to. It could have a constructor that handles the slots (if a slot isNULL
: call standard `dist``). Finally, this class could have a method that gives the distance matrix.Here is a little diagram to illustrate the concept.
The drawback of this idea is that the inheriting classes need to use the distance matrices. And I haven't found another class in bluster that uses it. At start, I thought this could help improve the maintainability of the code, but if we can't find other clustering method that use this, it may be of no use.
So as a sum up this class would be useful to any class that uses a distance matrix. The idea would be that these classes would inherit from the abstract class but maybe there are other possibilities.
What do you think?
The text was updated successfully, but these errors were encountered: