Deep Recurrent Q-Network (DRQN)

** Keras implementation of DRQN. This code needs to be fixed though its score would be good at the end. Please do not see this repository seriously.

Atari-Breakout

Main Idea?

CRNN model using 'TimeDistributed' wrapper for Conv2D and LSTM

def build_model(self):

    model = Sequential()

    model.add(TimeDistributed(Conv2D(32, (8, 8), strides=(4, 4), activation='relu'),

                              input_shape=(10, 84, 84, 1)))

                              #input_shape=(time_step, row, col, channels)
                              
    model.add(TimeDistributed(Conv2D(64, (4, 4), strides=(2, 2), activation='relu')))
   
    model.add(TimeDistributed(Conv2D(64, (3, 3), strides=(1, 1), activation='relu')))
   
    model.add(TimeDistributed(Flatten()))

    model.add(LSTM(512))

    model.add(Dense(128, activation='relu'))

    model.add(Dense(self.action_size))

    model.summary()

    return model

Breakout DRQN Source Code

Code Reference: 'breakout_dqn.py' at https://github.com/rlcode/reinforcement-learning-kr