Simple Reinforcement Learning with Tensorflow Part 0: Q-Learning with Tables and Neural Networks
Arthur Juliani

Hi Arthur,

Great Post!I am following this series of yours for quite some time now.I must say, it’s quite helpful.I am trying Q Learning on CartPole-v0 using tensorflow.Using exact same method that you showed here, the agent seems not to learn at all.Then I tried changing the loss function to Cross entropy and optimizer to AdamOptimizer.Still the issue remains.

However, Policy Gradient approach works for this.I am wondering if Q learning is right approach t0 the set ups like Cartpole?