Rsomoclu with negative values about somoclu

[peterwittek]

I just tried it from Python, and yes, codebook values can take negative values. Naturally, the initial codebook has a huge impact on the final map, but this is expected for any SOM implementation, so I do not find it necessary to elaborate it further in the documentation.

It is, however, not advised to have negative values in the features, especially if the codebook is initialized from 0 to 1. This comes from the batch training nature of the parallel implementation. The batch update rule will change the codebook values with weighted averages of the data points, and with negative values, the updates can cancel out. This is a good point, I have not thought about this before, and it is definitely worth documenting. Thanks for pointing this out, I will update the documentation when I can.

[peterwittek]

Commit dcb1b22 extends the documentation by a paragraph on this.