/* ---- Google Analytics Code Below */

Monday, July 25, 2016

Need for Dynamic Machine Learning

In one sense all typical, static machine learning methods are wrong.  That is they are based on past data, and the context that drove that data has changed.   The question is,  are the resulting models 'too wrong' to get useful results from?  Integrating dynamic (time varying) elements to the model can address this.  It does often add considerable complexity to the modeling effort.   A good introduction out of DSC on this topic.  

No comments: