Mobile Deep Learning with TensorFlow Lite, ML Kit and Flutter

Implementing the neural network

In this final section, we shall understand the neural network we have created for the agent to train with. We'll explore the util/nn.py file, which provides the NN class, along with the following important methods:

• __init__(self, filename): This function creates a new model using the util/keras_model.py functions if none exist on the disk. Otherwise, it loads the model file into the program.
• The model defined in the util/keras_model.py file is a residual CNN, which, in combination with MCTS and UCT, performs like a deep reinforcement learning neural network. The model formed has the following configurations:

input_dim: (2, util.BOARD_SIZE_H, util.BOARD_SIZE_W),policy_dim: util.KEY_SIZE,res_layer_num: 5,