07-14-2014 05:51 PM
I have a tree with the root node at the top. Each node value is a vector of length 100k. The goal is to find values of all nodes especially the parent nodes. The input are the leaf node vectors: if there are ten leaves, then there will be ten vectors of length 100k, or a 100k by 10 matrix . The output will be all the aggregated vectors: the value of any parent is the sum of all its children and the value of the leaf is simply the input. If there are 30 nodes in the tree, the output will be a 100k times 30 matrix. Does anyone know a data structure and algorithm to do this efficiently? If I would like to do this by SAS/IML, it there a way to do it? The tree can be any given tree. Thanks.
07-15-2014 11:53 AM
I don't know how to post data and how to remove personal information on attached file. The desired result can be simply a data table or a matrix provided you have the node name for each column.
07-15-2014 11:58 AM
At this moment, I don't need to add/delete or prune branches. Maybe in future I need. At present, assume that I have a fixed tree which is already constructed, we just need to aggregate tree at each height of tree only. I tried to upload a tree picture, but I failed.