Text this: Learning Policies for Neural Network Architecture Optimization Using Reinforcement Learning