- Windows 10 Home Edition
- msys64
- GCC 8.3.0
- CMake
- Eigen 3.3.7
- googletest 1.8.1
mkdir build && cd build
cmake ..
make
First, you need preparing the dataset consists of train, validation and test. The training data is needed to learn your model. It consist of labels and graph information. The validation data is needed to compare the performance for each optimizers. The testing data is used for computing outputs you want to know.
I assume your folder tree has become following structure.
dataset
|___ train
| |___ train
| |___ valid
|___ test
Each folder has graph data and label data (but test data doesn't have label data).
You can refer an example of input format of graph data in random_graph.txt
.
It consist of the number of vertecies and connections.
In train/valid dataset, you should name the graph and label file name with N_graph.txt
, N_label.txt
respectively (N is an id of data).
Each label file has one number corresponding to a graph label.
Move to build
folder and execute next command.
./GraphNN iteration batch_size dataset_path output_path
then, the results of optimization and the outputs with test data under results
folder.
make test
or
ctest