You need to install tensorflow, numpy and scipy
To install tensorflow and numpy, follow this guide: tensorflow
It is suggested to install the system on Virtualenv Virtualenv
To run the system, and if you are using Vritualenv, you have to execute the following command:
source path_to_virtualenv/bin/activate
Then you should run the FNN_learning.py file in FNN_python folder: Execute the following
python FNN_learning.py
To view the training data, run the following code: tensorboard --logdir=log path the log path is the one that includes the log files, in this case, it is dqn_log_files Currently, the cost function would be printed in the terminal, so checking on tensorboard is not necessary.
Models are saved using checkpoints functions in tensorflow APIs, the saved models are saved in: FNN_python/model
We need an 'unbiased' dataset to train the model and observe its performance