All Machine Learning Models have Flaws

Interesting view of alternative models to find models that describe data.   And that all models have flaws. This is a technical view, that is probably not very useful to the decision maker thinking about alternatives.  Useful for the technical practitioner as a reminder of the underlying issues.  This also does NOT include the most important of all flaws ... a model includes, or does not include, for any number of reasons, variables that may be important.  This is often the choice of the modeler, or their ignorance.  Often key in the real world.

(Update) And more here on this article from Data Science Central, via mention by Kirk Borne