Automatic Model Training

Is there a way modelling in rasa is done automatically, learning from past interactions with user? Perhaps with reinforcement learning?