somoclu
somoclu copied to clipboard
Massively parallel self-organizing maps: accelerate training on multicore CPUs, GPUs, and clusters
This problem is again back, which was removed in 1.7.5.1 version. Also, unfortunately now 1.7.5.1 version cannot be installed with pip. what could be the reason. Please guide.
The result after each iteration of som is the same, no change
I saw this issue https://github.com/peterwittek/somoclu/issues/80 but can't we still just simplify it without splitting into parts and concatenating over a loop? Or am I missing something?
I have probability input vectors and I would like to use the Jensen-Shannon Distance instead of the Euclidean one to train the map. Is it possible out of the box...