search

ARSadri / RobustGaussianFittingLibrary

Robust Gaussian Fitting Library
MIT License
5 stars 1 forks source link

get out of desert #6

Open ARSadri opened 4 years ago

ARSadri commented 4 years ago

support rkmeans

ARSadri commented 1 year ago

Theoretically, getting out of desert, means that we know where structures are. That is to make samples that are pure, because sampling is always possible.

What are the way to get out of desert? 1 - make samples small 2 - get the graph so far and sample from the graph. This is why HOSC worked, because they sampled from the old graph and iterations would help. 3 - If we have multiple models, they provide different sorting indices, very similar to the graph, can we make a sorting that benefits from all sortings?