> I will try to setup some as I did for SVR for anomalies data, and you will see it requires serious effort to train.

Sounds good to me. So far I've only got as far as seeing a few of the difficulties. I thought it might be useful to list a few of the different techniques currently in play.