DQN-Tensorflow

This is a simple implementation of DQN, based on tensorflow.
You can run run_this.py.
In this implementation, the reference codehttps://github.com/MorvanZhou/Reinforcement-learning-with-tensorflow/tree/master/contents/5_Deep_Q_Network
We set up a dual network structure for estimating the network and the target network, using the inconsistency of parameters between the two networks to reduce the correlation and calculate the loss function.

The input to the neural network is observation, the state value.
The output is the Q value of each action in this state.
The picture shows the change of loss during training.
file:///C:/Users/lenovo/Desktop/DQN_cost.png