Deep model is progressing much more rapidly than the statistical card-trick algorithms, Boltzmann machines, and in specific Restricted Boltzmann machines alread

Deep model is progressing much more rapidly than the statistical card-trick algorithms, Boltzmann machines, and in specific Restricted Boltzmann machines already vastly outperform SVM (By which I mean to say both Structured and Support vector machines) in a number of hidden model discovery tasks related to CV and is taking a lot of the existing 'state of the art' backprop classifiers (classical Neural Net & k-means clustering) to town. That said, traditional stochastic & markov driven approaches are more familiar and are cheaper to implement, which has hitherto driven the development of virtually all sciences.