what if FastICA doesn't converge?

[jankounchained]

this might be a more general problem with giving arguments to the underlying ICA algorithm.
for example:

• increase n_inter
• giving it a random state so you get exactly the same S3 model when you run it twice (nice for deployment)

[x-tabdeveloping]

I'm on it. Though I would argue that if the model doesn't converge after 200 iterations, it is probably a sign of something (there is no clear pattern in the data)

[jankounchained]

solved by #38