The MLPClassifier in sklearn.neural_network seems to use a lot of available CPU cores, i.e. the python process starts using 50% of processing power when fitting the model. How to prevent this? Is it possible? From the documentation it seem that there is no n_jobs parameter to control this behaviour. I am using sklearn version 0.20 from the November Winpython distribution. This is fine on my notebook but not good on a server others use, too. Continue reading...
