Smart Gov NN is used to learn best configurations of SmartGovLezModel. SmartGovLezModel, if used with Result Generator Task will create a file with configs and their results. These datas can be used to create a neural network in order to predict the results of new configurations.
To use this tool, you need to have :
- Python
- Tensorflow
- Keras
Once your file created by SmartGovLezModel, you have to train your neural network. You can also use the file config.txt that contains the results of around 86 000 simulations. To train your model :
python main.py -t <file>
The file in argument is the one with the configurations and the results. It will create a new file model.h5 in your repository, which contains your newly trained neural network.
Once your neural network is trained, you can use it to find the best configurations with a genetic algorithm :
python main.py -u <file>
The file in argument is the one containing your neural network (probably model.h5 if you did not rename it). It will load the neural network, and then launch a the research of best configurations using genetic algorithm for 60s. It will create a new json file in the JSON folder, with a configuration that can be visualised with a lez-viewer.