Hi Arun,

DQN doesn’t necessarily need to be composed of a convolutional layers, as would be required for a task with images as states. You could compose a DQN of fully-connected layers and that could be used to solve something like CartPole.

I have a tutorial where I show how to use PolicyGradient to quickly solve CartPole if you are interested: https://medium.com/@awjuliani/super-simple-reinforcement-learning-tutorial-part-2-ded33892c724#.t0020e9cu