Training a neural network using deep Q learning.

Before training

After training

Score

The mean for the last 100 episodes

The model

3 hidden layers with 64 neurons in each layer using relu as an activation function

The output layer contains 4 (Q value for each action) neurons

Usage

train the model

python train.py

test the model

Change my_model_1900.hdf5 to your model file name in test.py then run

python test.py