Self-organizing Vs k-mean cards

Question

Does anyone know how well self-organizing cards (SOMs) compare with k-tools? I find that usually in a color space such as RGB, SOM is the best way to combine colors together, since there is overlap in the color space between visually different colors ( http://www.ai-junkie.com/ann/som/som1. html ). Are there any cases where k means superior to SOM?

Thanks!

Answer 1

K-tool is a specialty of SOM, I believe. I am sure you can build ideal cases for him. I think that computational speed is the main advantage - when you gradually improve AI algorithms, sometimes more iterations of the worst algorithm give better performance than fewer iterations of the slower algorithm.

It all depends on the data. You will never know until you run it.

Answer 2

K-means a subset of self-organizing maps (SOM). K-means is a strictly average n-dimensional vector of n-spatial neighbors. SOM is similar, but the idea is to make the candidate vector closer to the corresponding vector and increase the difference with the surrounding vectors, disturbing them; perturbation decreases (core width) with distance; this is where the self-organizing part of the name comes from.

Answer 3

Self Organizing Maps creates two-dimensional output. k-means multidimensionality. SOMs operate in a discretized representation (grid). SOMs use a more local rule (neighborhood function). k-tools are more widely used as a clustering algorithm.