I have read this paper https://arxiv.org/pdf/1803.03674v1.pdf for outlier detection problem with real data (online training)
In this paper, the authors used Incremental Decision Tree to build subspaces (nodes in tree) by 2-mean method. The problem is in the algorithm 2 (build tree in paper), the condition (line 22): if t = beta power k. with t is the round observations (maybe each round we observe one sample) the splitting is performed.
The question is: What is beta and k? I spent a lot of time thinking about this. In the experiment you can see the authors plot chart describe splitting space process. At round 5, 11, 40? These are t (in my understanding) 5 can't be a result of any number power any number.