Michiel de Roo comments

Results 4 comments of


                                            Michiel de Roo

Implement self-organizing maps (SOMs) in `linfa-clustering`

Maybe also look at generative topographic mapping, which uses a specialised form of the gaussian mixture model and the EM algorithm to achieve the same objective. https://proceedings.neurips.cc/paper_files/paper/1996/file/4e4e53aa080247bc31d0eb4e7aeb07a0-Paper.pdf

Implement self-organizing maps (SOMs) in `linfa-clustering`

I don't think you can compare SOM to k-means because it serves more a dimensionality reduction purpose. Compared to PCA it projects data onto a non-linear plane instead of a...

Implement self-organizing maps (SOMs) in `linfa-clustering`

It is definitely not wrong to say SOM is clustering. But to me the dimensionality reduction properties is what sets it apart. I think a separate crate for SOM would...

Implement self-organizing maps (SOMs) in `linfa-clustering`

linfa-kohonen sounds good to me. @relf Is it an idea to start a linfa-contrib crate, where we can put the algorithms that don't fit in the core of linfa. Scikit...