How to apply the LBM with MNAR missing data on the data of French parliament 2018.

The use of at least one GPU is necessary to run the model on the French parliament dataset.

The French parliament 2018 dataset is availble in folder data_parliament

Installation

Pytorch installation

The model is implemented with pytorch. To install pytorch we refer the reader to the Pytorch website

With conda:

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

With pip:

pip install torch torchvision

Other requirements

With conda:

conda install numpy
conda install -c anaconda scipy
conda install -c conda-forge matplotlib
conda install -c conda-forge argparse

With pip:

pip install numpy scipy matplotlib argparse

Usage

To run the model on the dataset, use the script run_on_dataset_parliament.py:

python run_on_dataset_parliament.py

The default number of row classes is 3 and column classes is 5.

To run with a GPU use the argument device and specify the cuda index of desired gpu (often 0):

python run_on_dataset_parliament.py --device=0

To run with higher number of classes, use the arguments nb_row_classes and nb_col_classes as:

python run_on_dataset_parliament.py --nb_row_classes=3 --nb_col_classes=5

With higher number of classes, the memory of your GPU may overflow. In that case, you can use a second GPU with the argument device2 (index cuda needs to be specify):

python run_on_dataset_parliament.py --device=0 --device2=1 --nb_row_classes=3 --nb_col_classes=8

The script can be keyboard interrupted at any moment. In that case, the algorithm returns the MPs and texts classes and a plot of the voting matrix re-ordereded according to class memberships.

License

[MIT]