The SVD based Caratheodory reduction algorithm is shared across variations and can be implemented independent of the tree-based data structure. Feasibility and utility of the Tree-based data structures is to be determined.
See here for a tree-free implementation in numpy/python.
What's the new feature?
Sub task of #71
1-Tree (a.k.a. LittererLyons) recombination. Refs: Litterer & Lyons 2012 and Tchernychova 2015 Chapter 3.2 Algorithm 5The SVD based Caratheodory reduction algorithm is shared across variations and can be implemented independent of the tree-based data structure. Feasibility and utility of the Tree-based data structures is to be determined.
TODO:
What value does this add?
Provides additional algorithms for generating coresets (reduced measures)
Is there an alternative you've considered?
No response
Additional context
No response