(Scroll to middle) In many ways, machine learning is about pattern recognition and detecting those events that do not fit a particular pattern:  anomalies. Yet anomaly detection — a frequently touted capability of many systems — often does more bad than good, some data analysis experts argue.