RBM and Auto-encoder comparison


im trying to implement a recommender system using RBM and deep auto-encoder architectures separately.RBM try to predict ratings for not rated movies(1 like/0 no),and AE try to predict the exact rating of the movie(from 1 to 5 stars).
i used the RMSE and MAE metrics for comparison.however i noticed that they are note necessarily used for the same thing,RBM use binary inputs to predict features accuracy (%),and AE use inputs to predict the user rating for a specific movie . so we can not compare them using RMSE and MAE ?(in my case i get MAE of RBM 0.24 (75%accuracy) and with AE0 i get 0.77 (if user gives rate 4 stars we systeme can give between 3 and 4) not exactly the same job .what do u say?