A Minimalistic Approach to Sum-Product Network Learning for Real Applications

   Abstract